System for generating a video based multi-part response to a video input query using machine-learned models

Abstract

The present disclosure provides computer-implemented methods, systems, and devices for generating multipart explanations that include annotated video. A computing device accesses visual data depicting a scene and a query associated with the visual data. The computing device generates an input to a machine-learned model based on the visual data and the query. The computing device receive, as output from the machine-learned model, a multi-part response to the query, the multi-part response comprising a visual indication of the part to be performed with respect to the scene. The computing device provides the multi-part response for display.

Description

SYSTEM FOR GENERATING A VIDEO BASED MULTI-PART RESPONSE TO AVIDEO INPUT QUERY USING MACHINE-LEARNED MODELSPRIORITY CLAIM

[0001] This application claims priority to and the benefit of each of United StatesProvisional Patent Application Number 63 / 616,442, filed December 29, 2023. Applicant claims priority to and the benefit of this application and incorporates this application by reference in its entirety.FIELD

[0001] The present disclosure relates generally to providing explanations in response to user requests. More particularly, the present disclosure relates to using machine-learned models to respond to queries that include videos.BACKGROUND

[0002] Improvements in technology have enabled a variety of different services to be provided to users. For example, search services can respond to general user queries quickly and efficiently. However, for specific types of queries, a general search system may not provide the most optimal response.

[0003] For example, it can be tedious and time-consuming for a user to search for the specific information that will help the user solve a particular problem on which they are currently working. Additionally, the user may find explanations that are not sufficiently customized to their current situation. As a result, users may find the process of using search engines to find tutorials and explanations so difficult that they cease to use search engines to find help for their problems at all.SUMMARY

[0004] Aspects and advantages of embodiments of the present disclosure will be set forth in part in the following description, or can be learned from the description, or can be learned through practice of the embodiments.

[0005] One example aspect of the present disclosure is directed to a computing system. The system can include one or more processors and one or more non-transitory computer- readable media that collectively store instructions that, when executed by the one or more processors, cause the computing system to perform operations. The operations can include accessing, by a computing system with one or more processors, visual data depicting a sceneand a query associated with the visual data. The operations can further include generating, by the computing system, an input to a machine-learned model based on the visual data and the query. The operations can further include receiving, by the computing system as output from the machine-learned model, a multi-part response to the query, the multi-part response comprising a visual indication of the part to be performed with respect to the scene. The operations can further include providing, by the computing system, the multi-part response for display.

[0006] Another example aspect of the present disclosure is directed to one or more non- transitory computer-readable media that collectively store instructions that, when executed by one or more computing devices, cause the one or more computing devices to perform operations, the operations comprising accessing visual data depicting a scene and a query associated with the visual data. The operations further comprise generating input to a machine-learned model based on the visual data and the query. The operations further comprise receiving, as output from the machine-learned model, a multi-part response to the query, the multi-part response comprising a visual indication of the part to be performed with respect to the scene. The operations further comprise providing the multi-part response for display to the user.

[0007] Another example aspect of the present disclosure is directed to a computing system, the system comprising: one or more processors; and one or more non-transitory computer-readable media that collectively store instructions that, when executed by the one or more processors, cause the computing system to perform operations. The operations can include receiving a query from a user, wherein the query includes a video. The operations further comprise generating input to a machine-learned model, the input including the video and a query request based on the query. The operations further comprise receiving, as output from the machine-learned model, a multi-part response, wherein the multi-part response provides instructions associated with performing one or more actions and at least one part in the multi-part response includes text data and visual data. The operations further comprise providing the multi-part response for display to a user.

[0008] Other aspects of the present disclosure are directed to various systems, apparatuses, non-transitory computer-readable media, user interfaces, and electronic devices.

[0009] These and other features, aspects, and advantages of various embodiments of the present disclosure will become better understood with reference to the following description and appended claims. The accompanying drawings, which are incorporated in and constitutea part of this specification, illustrate example embodiments of the present disclosure and, together with the description, serve to explain the related principles.BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Detailed discussion of embodiments directed to one of ordinary skill in the art is set forth in the specification, which makes reference to the appended figures, in which:

[0011] FIG. 1 depicts a block diagram of an example computing system that generates multi-part response to video-based queries according to example embodiments of the present disclosure;

[0012] FIG. 2A depicts an illustration of an example interface for displaying a video captured by the camera of a user computing device accepting according to example embodiments of the present disclosure;

[0013] FIG. 2B depicts an illustration of an example interface for receiving a query using voice input according to example embodiments of the present disclosure;

[0014] FIG. 2C depicts an illustration of an example interface for displaying a multi-part explanation for completing the task according to example embodiments of the present disclosure;

[0015] FIG. 2D depicts an illustration of an example interface for displaying a multi -part explanation for completing the task according to example embodiments of the present disclosure;

[0016] FIG. 3 A illustrates an example interface for displaying a video captured by the camera of a user computing device accepting according to example embodiments of the present disclosure;

[0017] FIG. 3B illustrates an example interface for receiving a query from a user according to example embodiments of the present disclosure;

[0018] FIG. 3C illustrates an example interface for displaying an explanation for performing an exercise properly according to example embodiments of the present disclosure;

[0019] FIG. 3D illustrates an example interface for displaying a multi-part explanation for completing the exercise according to example embodiments of the present disclosure;

[0020] FIG. 4A illustrates an example interface for displaying a video captured by the camera of a user computing device accepting according to example embodiments of the present disclosure;

[0021] FIG. 4B illustrates an example interface for receiving a query from a user according to example embodiments of the present disclosure;

[0022] FIG. 4C illustrates an example interface for displaying an explanation for performing an exercise properly according to example embodiments of the present disclosure;

[0023] FIG. 4D illustrates an example interface for displaying a multi-part explanation for completing the exercise according to example embodiments of the present disclosure;

[0024] FIG. 4E illustrates an example interface for displaying a multi-part explanation for completing the exercise according to example embodiments of the present disclosure;

[0025] FIG. 5A depicts an example interface for application that presents visual data captured by a camera of the user computing device and analyzes the contents of the visual data to provide services to the user in accordance with example embodiments of the present disclosure;

[0026] FIG. 5B includes an example interface that can display a multi-part explanation to the user in response to the query request submitted by the user in accordance with example embodiments here;

[0027] FIG. 6 represents an example system for providing multi-part response to queries that include visual data according to example embodiments of the present disclosure;

[0028] FIG. 7 depicts a block diagram of an example computing system that performs a method for generating multi-part explanations that includes visual data according to example embodiments of the present disclosure;

[0029] FIG. 8 depicts a block diagram of an example computing system that performs a process for generating multi-part responses to video queries according to example embodiments of the present disclosure;

[0030] FIG. 9 depicts an example flow diagram for a method of generating a multi-part response that include annotated video data according to example embodiments of the present disclosure; and

[0031] FIG. 10 depicts an example flow diagram for a method of generating multi -part responses to videos queries according to example embodiments of the present disclosure.

[0032] Reference numerals that are repeated across plural figures are intended to identify the same features in various implementations.DETAILED DESCRIPTION

[0033] Generally, the present disclosure is directed to systems and methods for providing explanations in response to video queries (e.g., a user query that includes a video aspart of the query) using machine-learned models. In particular, the systems and methods disclosed herein can provide information for receiving a user query that includes visual data. The visual data can be a video or one or more images that depict a scene. The query can also include query data. A query response system can generate input to a multimodal machine- learned model based on the visual data and the query. The multimodal machine-learned model can generate a response to the input as output. The response can include a plurality of parts (e.g., part can be a step in a multistep process). The response can include an annotation on an image that represents a visual indication of at least one part being performed with respect to the same scene as depicted in the visual data.

[0034] In some examples, the visual data is a video depicting the performance of a particular task. In some examples, the query response system can infer the query from the task pictured in the video or the audio associated with the submitted video. The query response system can generate an input and provide it to the multimodal machine-learned system. The multimodal machine-learned model can output a response. The response could include parts or instructions for performing the task depicted in the input video. In some examples, the response can include a visual depiction (e.g., an image or video) based on the input video or image. For example, the output can include the input video along with annotations that explain how a particular process is to be performed or a particular goal is to be achieved. In other examples, the response can include instructions for performing the task without a specific depiction of the original scene.

[0035] For example, a user can be in the process of changing the air filter in their car. During this process, the user can capture video of their vehicle’s engine with a camera included in their user computing device (e.g., a smartphone). The user can provide the query, “How do I change the air filter in my car?” The video and the associated query text can be provided to the query response system. The query response system can receive the video and the accompanying query text. It can generate a prompt for the multimodal machine-learned model. The prompt can be provided to the multimodal machine-learned model, which can output a multi-part response. The machine-learned model is multimodal in that it can use textual data, visual data, and audio data as input. Similarly, it can provide output consisting of textual, visual, and audio data. The multi-part response can include a series of parts for changing the air filter in the vehicle. At least one step in the series of steps can include the visual scene that was depicted in the original video with annotations or other modifications that make the step being performed easier to understand. For example, suppose a particular step includes removing a bolt. In that case, that step can have an associated image extractedfrom the video. The extracted image can be annotated to highlight the specific bolt that needs to be removed. In this way, the instructions provided by the machine-learned model can be easily understood with respect to the specific situation that the user is dealing with.

[0036] More generally, a query response system can be enabled by an application on a user computing device, by communicating with a remote server system, or a combination of both. A server computing system can be any computing system configured to communicate with a user computing device (or other computing devices) over a network to provide information or a service. If a server computing system is employed, the server computing system can receive, from a user computing device, queries, requests, videos (or other visual data), and so on. The user computing device can provide queries, requests, and videos through an application.

[0037] A user computing device can be any computing device designed to be operated by an end-user. For example, a user computing device can include but is not limited to a personal computer, a smartphone, a smartwatch, a fitness band, a tablet computer, a laptop computer, a hand-held navigation computing device, a wearable computing device, a game console, and so on. In some examples, a user computing device can include one or more sensors intended to gather information, with the permission of the user, such as an image sensor (e.g., a camera).

[0038] In some examples, a query response system can receive a request from a user. The request can be received in a variety of formats. In some examples, the request can be received as input while the user operates an application. The application can display captured video data and provide some live analysis of that video data. The application can include a query input field in which a user can enter a query request. For example, a user may type a particular query into the search input field associated with the application. The user can also designate a portion of the video (e.g., a previously captured video segment or image) to be associated with the query request.

[0039] Alternatively, the user can generate a video after the query is entered using the application and the camera associated with the user computing device to capture video for context when generating a response to the query. In other examples, the application can analyze a video currently being captured to determine one or more applicable queries and provide a prompt (via a user interface element) suggesting a response to the one or more applicable queries. For example, if the user is capturing a video of a meal in a restaurant and the bill arrives in the video data, the application may offer to calculate an appropriate tip.

[0040] In some examples, a server computing system can receive the query and accompanying video data via a communication network. The query can include query text indicating the specific request from a user.

[0041] Once the query has been received, the query response system can generate input to a machine-learned model. The input can include a prompt that includes, but is not limited to, the query text associated with the query, the visual data, other contextual information, and instructions for the format of the output of the machine-learned model.

[0042] In some examples, the one or more machine-learned models can be any of a variety of various machine-learned models such as neural networks (e.g., deep neural networks or large language models), other types of machine-learned models, including nonlinear models and / or linear models, or binary classifiers. Neural networks can include feedforward neural networks, recurrent neural networks (e.g., long short-term memory recurrent neural networks), convolutional neural networks, Transformer networks, denoising diffusion models, or other forms of neural networks.

[0043] In some examples, multiple machine-learned models can be used to improve the usefulness of explanations. A first machine-learned model can be a model that generates multi-part explanations based on the submitted video and the query text. This machine- learned model can be a large language model. The first machine-learned model can take a video and query text as input and outputs a multi-part explanation that corresponds to the input video as output.

[0044] In some examples, a second machine-learned model can be used to generate appropriate videos and / or images as needed by the first machine-learned model. The first machine-learned model can output a series of steps to perform a task or correct an error. Each part may include detailed explanations for what to display in the user interface to the user, including text, images, videos, and the layout / formatting used when displaying the multi-part response. In some examples, the content to be displayed can include a portion of the video query, including either the video itself, a portion of it, or a single image from the video. The portion of the video to be displayed at a particular step in the multi-step process can include visual annotations that explain the particular step with respect to the scene displayed in the query video.

[0045] In some examples, the annotation can include highlighting a portion of the image or video, demonstrating an action to be taken, showing a correction to a portion of the video, and so on. For example, suppose the query video depicts a user performing a deadlift, and the query is a request to improve the form of the user performing the deadlift. In that case, theannotations can include visual indications of changes to the user’s posture and movement that will improve their form. For example, the annotation could include a visual indication that the weight bar should be held closer to the user’s legs and that the user should straighten their back during the lift.

[0046] In some examples, the output of the first machine-learned model can provide instructions that describe how and where the annotations should be placed. In some examples, the output for the machine-learned model can also include text. The text can be formatted to include a series of steps to be performed. Each part or step can have associated visual content that can include one or more portions of the video query that have been annotated to represent the content of the respective part or step.

[0047] The query response system can provide input to the first machine-learned model based on the query and the associated video. In some examples, the input to a machine- learned model, especially if the model is a large language model, takes the form of a prompt. The prompt can be a specific request of the large language model and sets the parameters for the response the large language model will give to the prompt. The prompt can include the query and provide the associated visual data.

[0048] The query response system can also access relevant contextual information to include with the query data and video. For example, the contextual information can include similar or related queries previously received by the query response system. This can include the query, the content extracted from the video or the video itself, the generated prompt, and the output produced by the multimodal machine-learned model. In some examples, the query response system can also perform a search of the web (or other store of data) and provide the results as context in a prompt.

[0049] Contextual information can include information about the user that the user has agreed to provide as part of receiving the appropriate explanation for the submitted query or problem. For example, the prompt can include the current understanding level of the user. For example, a user with a basic understanding of a topic may receive a more in-depth explanation of particular steps than would be required for a user with a more significant knowledge of the subject.

[0050] The prompt can also include topic-specific information associated with the query or problem. For example, if the problem to be solved is associated with a particular topic, the query response system can retrieve information about that topic from a database and provide it to the multimodal machine-learned model as part of the input (e.g., as additional “context” input). In some cases, providing additional background information can be useful in ensuringthat the response provided by the multimodal machine-learned model is as accurate as possible.

[0051] In some examples, the request associated with a particular request can be part of an ongoing interaction with a user. For example, the user could ask a series of questions related to one or more problems. The prompt can include the past conversation history or background associated with any particular request. In some examples, the user may ask a clarifying question about a previously presented explanation. Any past portions of the interaction (e.g., previous queries and their associated responses) can be provided as context for the clarification question.

[0052] In some examples, the multimodal machine-learned model can provide an output in response to the input. The output can be a multi-part explanation of how to accomplish a task (or solve a particular problem). A multi-part explanation is an explanation that includes a plurality of parts that need to be followed to accomplish the task.

[0053] In some examples, the output generated by the multimodal machine-learned model can be specific to a particular user. For example, the multi-part response can be tailored to the user’s understanding or needs. To do so, information included in a user profile (e.g., information about the user’s current understanding or mastery level for a topic) can be provided as context for the prompt, and the multimodal machine-learned model can generate a multi-part response that is directed towards the current level of understanding available to the user. The output data can be formatted such that each part in the multi-part process is displayed in a particular portion of the user interface. As discussed above, each part can include text, visual data associated with the video (e.g., portions of the video or images from the video), annotations to the video, and so on. In some examples, the parts are displayed in separate sections of the interface based on formatting data (e.g., markup data) included in the output or otherwise provided by the machine-learned model. In some examples, each section can be collapsible. In this way, the collapsed version of a part may include a basic statement of what the part is. The user can then select an interface element that causes the section to expand. The expanded section can display a significantly more detailed version of that part (e.g., a particular step in a multistep process), including at least one portion of the video with annotations overlaid on the image.

[0054] In this way, users can determine which parts they want to view in detail and which parts can be represented by a brief summary (e.g., the user only expands parts they are not already familiar with). Collapsible sections can allow users to decide which parts are displayed in detail, resulting in an easier-to-understand and less crowded user interface.

[0055] In some examples, the output of the machine-learned model designates one or more of the parts as being initially collapsed and one or more of the parts as initially expanded. In some examples, the output of the multimodal machine-learned model can designate particular parts as being expanded or collapsed based on information about the current level of understanding that the user currently has (e.g., context data input into the machine-learned model in a prompt). Thus, for some advanced users, the multimodal machine-learned model may designate the complicated parts to be expanded and the simple parts to be collapsed. In contrast, when generating a response for a beginning user, the multimodal machine-learned model may determine that the first part should be expanded with the expectation the user will expand each part as they go through the process described in the multi-part process.

[0056] In other examples, one or more parts are initially expanded based on their position in the part order (e.g., the first part is expanded) or the difficulty of the content in a particular part (e.g., the most difficult parts are expanded and more straightforward parts default to being collapsed).

[0057] Once the multimodal machine-learned model provides the output, the output can be displayed to a user. In some examples, if the machine-learned model is provided at a server computing system, the data will be transmitted via the computing network to the user computing device. In other examples, the machine-learned model is executed by the user computing device and does not need to be transmitted via the network.

[0058] In some examples, the output of the multimodal machine-learned model includes a confidence value, the confidence value representing the degree to which the multimodal machine-learned model is confident in the validity of the generated multi-part response. In some examples, the response can be displayed as usual if the confidence value is above a predetermined threshold. However, if the confidence value is below the threshold, the query response system can provide a less detailed explanation (a hint) that avoids factual inaccuracies.

[0059] For example, if the video query includes a video of a user playing a video game and the query text “How do I solve the water temple puzzle in Maze Solver 12?”, the multimodal machine-learned model can determine that the proposed response may not be reliable based on a confidence value. Instead of displaying a response with a low confidence value, the query response system can display a briefer answer, such as “I’m not sure how to solve this, but here is a list of guides for solving water temple puzzles in video games.” Inthis way, the user can determine whether the explanation is sufficient and, if not, provide more detail to the system.

[0060] In some examples, the video submitted with the query can be analyzed to determine an answer to the query. For example, the user can submit a video of a user on a trip and request the query response system to generate an itinerary based on the trip in the video. The query response system can, using a machine-learned model, analyze the video to determine a plurality of video segments in the video. Each video segment can be associated with one or more features. The features can be associated with a particular semantic topic or meaning. In some examples, the feature can be an object pictured in the video. For example, all the portions of a video depicting object A can be grouped into a segment, and all the portions of the video depicting object B can be grouped into a different segment.

[0061] Segments can be grouped based on a particular part within a process that is being performed or a particular location (or a location type) is being displayed. If the video includes actors, the segments can be generated based on the people that are pictured. Once the video has been grouped into a plurality of segments, the query response system (or an associated machine-learned model) can determine a multi-part response to the query. One or more segments of the video can be associated with a part in the multi-part response.

[0062] For example, if the video depicts a recipe being made, the query response system can analyze the video to generate a plurality of distinct segments in the video. The distinct segments can represent each step or ingredient of the recipe. Based on the video or another source of supplemental information, the query response system can determine the recipe and generate a multi-part explanation describing how to make the recipe. For each step of the recipe, the query response system can determine whether there is a corresponding segment of the video. If so, the information associated with the part can be included in the information presented to the user about the part.

[0063] In some examples, the output of the multimodal machine-learned model includes a video. The video can be an explanation of how to perform a task described in the multi -part response. The query response system can determine a plurality of segments within the video. The query response system can associate one or more segments within the video to one or more parts of the multi-part response. In some examples, the output of the multimodal machine-learned model can describe the different video segments and detail the specific segments of the video to be associated with each part.

[0064] The systems and methods of the present disclosure provide a number of technical effects and benefits. In one example, the system and methods can provide real-time,personalized explanations, solutions, and guides for various queries with associated videos submitted to the system. In particular, the systems and methods disclosed herein can receive a query with associated visual data, generate input to a multimodal model, and receive a structured multi-part explanation from the machine-learned model, including an annotated video or image from the query video for display to the requesting user.

[0065] Furthermore, providing a structured, multi-part response allows users to selectively decide what portions of the responses to review in detail. Doing so reduces clutter in a display, while simultaneously reducing the amount of memory used and bandwidth required by only transmitting detail when requested by the user.

[0066] Aspects of the proposed systems and methods also represent a technical solution to the technical problem of integrating multimodal content. The system can generate explanations that include images from the originally submitted video illustrating how to perform the specific part in the scene the user submitted. Doing so increases the likelihood that a particular user will understand the part correctly and perform it accurately. This increase in intelligibility reduces the cost of performing the task in terms of time and processing power and increases user satisfaction.

[0067] With reference now to the Figures, example embodiments of the present disclosure will be discussed in further detail.

[0068] FIG. 1 depicts a block diagram of an example computing system 100 that generates a multi-part response to video-based queries according to example embodiments of the present disclosure. The computing system 100 includes a user computing device 102, a server computing system 130, and a training computing system 150 that are communicatively coupled over a network 180.

[0069] The user computing device 102 can be any type of computing device, such as, for example, a personal computing device (e.g., laptop or desktop), a mobile computing device (e.g., smartphone or tablet), a gaming console or controller, a wearable computing device, an embedded computing device, or any other type of computing device.

[0070] The user computing device 102 includes one or more processors 112 and a memory 114. The one or more processors 112 can be any suitable processing device (e.g., a processor core, a microprocessor, an ASIC, an FPGA, a controller, a microcontroller, etc.) and can be one processor or a plurality of processors that are operatively connected. The memory 114 can include one or more non -transitory computer-readable storage mediums, such as RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., andcombinations thereof. The memory 114 can store data 116 and instructions 118 which are executed by the processor 112 to cause the user computing device 102 to perform operations.

[0071] In some implementations, the user computing device 102 can store or include one or more machine-learned models 120 (e.g., one or more multimodal machine-learned models). For example, the machine-learned models 120 can be or can otherwise include various machine-learned models such as neural networks (e.g., deep neural networks) or other types of machine-learned models, including non-linear models and / or linear models. Neural networks can include feed-forward neural networks, recurrent neural networks (e.g., long short-term memory recurrent neural networks), convolutional neural networks or other forms of neural networks. Example machine-learned models 120 are discussed with reference to FIGS. 7-8.

[0072] In some implementations, the one or more machine-learned models 120 can be received from the server computing system 130 over network 180, stored in the user computing device memory 114, and then used or otherwise implemented by the one or more processors 112. In some implementations, the user computing device 102 can implement multiple parallel instances of a single machine-learned model 120.

[0073] More particularly, the user computing device 102 (e.g., a model and / or a synthesis model) can receive a query from a user that includes visual data. The visual data can be a video, or a series of images captured by the user computing device 102. In some examples, the query can also include query text. The query text can be received from the user entry in a query text field. The query text can be extracted from audio provided by a user. In some examples, the query text can be generated based on a user interface element selected by a user. For example, if the user is capturing video of the user playing a level of a video game, the user interface can be updated (based on the detection of the level and the video game) with a user interface including a prompt offering an explanation in progressing through the level. If the user selects the user interface element with the prompt, the system can automatically generate a query associated with receiving an explanation for the particular game level, game area, or game challenge.

[0074] In some examples, the query text can be extracted from the audio associated with a video. For example, the video can include a user asking a question. When that video is received, the user computing device 102 can analyze the video to extract the audio query from the video.

[0075] In some examples, the visual data can be a video or one or more images submitted by a user. The video or one or more images can be captured by a camera associatedwith a user computing device and can represent a particular scene. For example, if the question is associated with repairing an appliance, the video can depict the specific appliance. In some examples, the visual data can also depict a user performing an action or process.

[0076] Once the query and video have been received, the user computing device 102 can generate a prompt as input to the multimodal machine-learned model. The prompt can include the text of the query, one or more portions of the video, and any contextual data that is necessary.

[0077] The multimodal machine-learned model can generate a response based on the input prompt. In some examples, the output can be a multi-part response, each part representing a part of an explanation explaining how to perform a step or portion of a particular task or solve a specific problem. In some examples, one or more parts can have associated visual data. The associated visual data can include the same scene as the original video submitted with one or more annotations. The annotations can be visual indications overlaid on images or portions of video from the original video.

[0078] In other examples, the annotations can include text that indicates information about performing the particular part with which the visual data is associated. In some examples, the visual data included with the part consists of the same visual data as the submitted video with alterations or embedded instructions for performing the associated part or task. In this way, the parts to be taken can be more easily understood with respect to the actual situation the user is in.

[0079] In some examples, the output can be structured in a multi-part format. For example, the output can be formatted such that each part can be assigned a particular collapsible element of the user interface in which it is displayed. Each part can be collapsible or expandable, so more or fewer details about a particular part can be shown depending on the user's preference. In this way, parts that the user already understands can be minimized, while parts that are difficult for the user can be expanded so the user has more information on how to complete that part.

[0080] The parts can be presented to a user in the user interface. In some examples, some parts can be automatically expanded. The automatically expanded parts can include the first, most complex, or most important parts. In some examples, the part that includes visual data with annotations can be initially expanded.

[0081] In another example, the user can receive an explanation request. The explanation request can be a specific request for detailed instructions for performing a particular task. In some examples, the explanation request can include a video, the contents of which depict theuser performing a task. The explanation request can also include a query that indicates the problem the user is having or the problem for which the user needs an explanation. For example, suppose the user has a question about performing a particular action or solving a particular puzzle in a video game. In that case, the video can include the user playing the portion of the video game associated with the problem or puzzle.

[0082] As mentioned above, the user computing device 102 can generate input to a machine-learned model in the form of a prompt. The machine-learned model can generate a multi-part response based on that input. The output of the machine-learned model can include an explanation showing how to perform an action, solve a problem, beat a boss, or overcome a puzzle. However, the output may not include image content from the original video.Instead, the output can include video data from other sources.

[0083] Additionally, or alternatively, one or more machine-learned models 140 (e.g., one or more multi-part response models and / or synthesis models) can be included in or otherwise stored and implemented by the server computing system 130 that communicates with the user computing device 102 according to a client-server relationship. For example, the one or more machine-learned models 140 can be implemented by the server computing system 130 as a portion of a web service (e.g., as part of a query response service). Thus, one or more models 120 can be stored and implemented at the user computing device 102 and / or one or more models 140 can be stored and implemented at the server computing system 130.

[0084] The user computing device 102 can also include one or more user input component 122 that receives user input. For example, the user input component 122 can be a touch-sensitive component (e.g., a touch-sensitive display screen or a touchpad) that is sensitive to the touch of a user input object (e.g., a finger or a stylus). The touch-sensitive component can serve to implement a virtual keyboard. Other example user input components include a microphone, a traditional keyboard, or other means by which a user can provide user input.

[0085] The server computing system 130 includes one or more processors 132 and a memory 134. The one or more processors 132 can be any suitable processing device (e.g., a processor core, a microprocessor, an ASIC, an FPGA, a controller, a microcontroller, etc.) and can be one processor or a plurality of processors that are operatively connected. The memory 134 can include one or more non-transitory computer-readable storage mediums, such as RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. The memory 134 can store data 136 and instructions 138 which areexecuted by the processor 132 to cause the server computing system 130 to perform operations.

[0086] In some implementations, the server computing system 130 includes or is otherwise implemented by one or more server computing devices. In instances in which the server computing system 130 includes plural server computing devices, such server computing devices can operate according to sequential computing architectures, parallel computing architectures, or some combination thereof.

[0087] As described above, the server computing system 130 can store or otherwise include one or more machine-learned models 140 (e.g., one or more multimodal query response models and / or synthesis models). For example, the models 140 can be or can otherwise include various machine-learned models. Example machine-learned models include neural networks or other multi-layer non-linear models. Example neural networks include feed forward neural networks, deep neural networks, recurrent neural networks, and convolutional neural networks. Example models 140 are discussed with reference to FIGS. 7- 8.

[0088] The user computing device 102 and / or the server computing system 130 can train the models 120 and / or 140 via interaction with the training computing system 150 that is communicatively coupled over the network 180. The training computing system 150 can be separate from the server computing system 130 or can be a portion of the server computing system 130.

[0089] The training computing system 150 includes one or more processors 152 and a memory 154. The one or more processors 152 can be any suitable processing device (e.g., a processor core, a microprocessor, an ASIC, an FPGA, a controller, a microcontroller, etc.) and can be one processor or a plurality of processors that are operatively connected. The memory 154 can include one or more non-transitory computer-readable storage mediums, such as RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. The memory 154 can store data 156 and instructions 158 which are executed by the processor 152 to cause the training computing system 150 to perform operations. In some implementations, the training computing system 150 includes or is otherwise implemented by one or more server computing devices.

[0090] The training computing system 150 can include a model trainer 160 that trains the machine-learned models 120 and / or 140 stored at the user computing device 102 and / or the server computing system 130 using various training or learning techniques, such as, for example, backwards propagation of errors. For example, a loss function can bebackpropagated through the model(s) to update one or more parameters of the model(s) (e.g., based on a gradient of the loss function). Various loss functions can be used such as mean squared error, likelihood loss, cross entropy loss, hinge loss, and / or various other loss functions. Gradient descent techniques can be used to iteratively update the parameters over a number of training iterations.

[0091] In some implementations, performing backwards propagation of errors can include performing truncated backpropagation through time. The model trainer 160 can perform a number of generalization techniques (e.g., weight decays, dropouts, etc.) to improve the generalization capability of the models being trained.

[0092] In particular, the model trainer 160 can train the query response models 120 and / or 140 based on a set of training data 162. The training data 162 can include, for example, example video queries with accompanying structured, multi-part explanations (or other ground truth data), training context data, and / or training motion data.

[0093] The model trainer 160 includes computer logic utilized to provide desired functionality. The model trainer 160 can be implemented in hardware, firmware, and / or software controlling a general-purpose processor. For example, in some implementations, the model trainer 160 includes program files stored on a storage device, loaded into a memory and executed by one or more processors. In other implementations, the model trainer 160 includes one or more sets of computer-executable instructions that are stored in a tangible computer-readable storage medium such as RAM hard disk or optical or magnetic media.

[0094] The network 180 can be any type of communications network, such as a local area network (e.g., intranet), wide area network (e.g., Internet), or some combination thereof and can include any number of wired or wireless links. In general, communication over the network 180 can be carried via any type of wired and / or wireless connection, using a wide variety of communication protocols (e.g., TCP / IP, HTTP, SMTP, FTP), encodings or formats (e.g., HTML, XML), and / or protection schemes (e.g., VPN, secure HTTP, SSL).

[0095] The machine-learned models described in this specification may be used in a variety of tasks, applications, and / or use cases. In some implementations, the input to the machine-learned model(s) of the present disclosure can be visual data. The machine-learned model(s) can process the visual data to generate an output. For example, the machine-learned model(s) can process the visual data to generate a multi-part response by extracting a query from the visual data.

[0096] In some implementations, the input to the machine-learned model(s) of the present disclosure can be text, natural language data, video or image data, and so on. Themachine-learned model(s) can process the t text, natural language data, video, or image data to generate an output. In some implementations, the input to the machine-learned model(s) of the present disclosure can be speech data. The machine-learned model(s) can process the speech data to generate an output. As an example, the machine-learned model(s) can process the speech data to generate a speech recognition output. The output of the speech recognition system can be used as input to the model.

[0097] FIG. 1 illustrates one example computing system that can be used to implement the present disclosure. Other computing systems can be used as well. For example, in some implementations, the user computing device 102 can include the model trainer 160 and the training dataset 162. In such implementations, the models 120 can be both trained and used locally at the user computing device 102. In some of such implementations, the user computing device 102 can implement the model trainer 160 to personalize the models 120 based on user-specific data.

[0098] FIG. 2A depicts an illustration of an example interface for displaying a video captured by the camera of a user computing device according to example embodiments of the present disclosure. In some implementations, a query response system can include one or more machine-learned models (e.g., a multi-part response machine-learned model or image generation machine-learned model) trained to generate multi-part explanations for performing a task or solving a problem.

[0099] A video display and analysis application can be used to capture video of a user performing a specific task. In this example, the user video display and analysis application can capture video of the user playing a video game. The camera of the user computing device 102 can be directed towards a screen on which the video game is being displayed to the user. The user interface of the video display and analysis application can include a query field 202, into which a user can input a query with respect to the displayed video.

[0100] The text of the query can be input into query field 202, and the service or application 200 can also capture a portion of the video for use in providing a response. In some examples, the user interface 204 of the website or application can include a user interface element for submitting the query or request.

[0101] In some examples, the query response system can receive the query and one or more portions of the video displayed in the user interface. The query and video can be used as input to a multimodal machine-learned model to generate a multi-part response for performing a task as indicated by the query text and the video. In this example, the query text recites, "How do I find the coconut cave?"

[0102] FIG. 2B depicts an illustration of an example interface for receiving a query using voice input according to example embodiments of the present disclosure. In this example, a user may, while displaying video captured by the camera of their user computing device 102, use vocal input to submit a query. In some examples, vocal input can be initiated by interacting with an element of the user interface display (e.g., pushing a voice input button) or by initiating vocal input using a vocal command such as "OK Google."

[0103] Once the user has initiated vocal input, the user computing device or an application installed on the user computing device can transcribe the vocal input and provide it to a query response system. The query response system can also take, as input, the currently displayed video or image as well as one or more portions of previously displayed video. The query and the video can be used as input to generate a prompt for the multimodal machine- learned model, and the multimodal machine-learned model can generate a response to the query. In this example, the query is associated with identifying the location of the coconut cave. The response can take into account the current position and progress of the user based on the video that was submitted along with the query.

[0104] In this example, a portion of the response can be displayed in a section of the user interface below where the query was input. The response can include a textual answer to the response as well as one or more links to locations to find more information.

[0105] FIG. 2C depicts an illustration of an example interface for displaying a multi-part explanation for completing the task according to example embodiments of the present disclosure. In this example, the interface includes a general response 232 to the query. The general response can include a description of the general solution to the task. Below the general response 232 can be a list of parts to perform the task. Each part 236 in the multi-part list can be displayed in a distinct portion of the user interface.

[0106] Each portion of the list of parts can be expandable. When in collapsed form, the part can include a brief description of the task as well as a link 238 to a video showing performance of the part (e.g., a particular step in a multi-step process). After the list of parts, the interface can also include an input field for asking follow-up questions. For example, if a user has a question about one of the parts or the general response, they enter their query in the input field 240 and the query response system can generate a response. In some examples, the previous query and response can be used as input for context such that the response to the query can be as topical and relevant as possible.

[0107] FIG. 2D depicts an illustration of an example interface for displaying a multi -part explanation for completing the task according to example embodiments of the presentdisclosure. In this example, the interface includes a general response 232 to the query. The general response can include a description of the general solution to the task. Below the general response 232 can be a list of parts to perform the task. Each part 236 in the multi-part list can be displayed in a distinct portion of the user interface.

[0108] Each portion of the list of parts can be expandable. In this example, the first part in the list of parts has been expanded. For example, the user has clicked on a caret included in the expandible user interface element. In response, the portion of the user interface associated with the part expands, displaying additional information (detailed explanation 239) about that part. The additional data can include more information about the linked video. After the list of parts, the interface can also include an input field for asking follow-up questions. For example, if a user has a question about one of the parts or the general response, they can enter their follow-up query in the input field 240. The query response system can receive the new query and provide a relevant response.

[0109] FIG. 3 A illustrates an example interface for displaying a video captured by the camera of a user computing device accepting according to example embodiments of the present disclosure. In some implementations, a query response system can include one or more machine-learned models (e.g., a multi-part response machine-learned model or image generation machine-learned model) trained to generate multi-part explanations for performing a task or solving a problem.

[0110] A video display application can be used to capture video of a user performing a specific task. In this example, the user video display and analysis application can capture video of the user performing an exercise. The camera of the user computing device 102 can be directed towards the user while the user is performing the exercise. The user interface 304 of the video display application can include a video display area 306 and a query input button 302 that is user-selectable.

[0111] FIG. 3B depicts illustrates an example interface for receiving a query from a user according to example embodiments of the present disclosure. In this example, a user may, while an application is displaying video captured by the camera of their user computing device 102, use a search input interface element (e.g., input element 306 in FIG. 3A) to cause the user interface to display a query input field 316.

[0112] The user can input query text into the user input field 316. The query text can be input using a virtual keyboard displayed in the user interface. Alternatively, the user can provide voice input, input using a wireless keyboard, or any other form of input. The query response system can accept the query text and capture at least a portion of the video displayedwhile the query text is being input. In some examples, the user can indicate which portion of the displayed video is associated with the query text.

[0113] The query response system can use the query text and the associated video to generate an input prompt for the multimodal machine-learned model. The prompt can also include relevant supplemental information. For example, the query response system can analyze the video to determine a topic and provide information about that topic as part of the input to the machine-learned model. In this example, the query text reads "what's the proper form for this exercise?" The query text can be accompanied by a video showing the performance of a particular weightlifting exercise. The query response system can provide contextual information about weightlifting (including information about the particular weightlifting exercise) to the multimodal machine-learned model.

[0114] FIG. 3C illustrates an example interface for displaying an explanation for performing an exercise properly according to example embodiments of the present disclosure. In this example, the interface includes a general response 322 to the query. The general response can include at least one part in a multi-part explanation describing how to perform a particular exercise properly. In this example, the exercise is a shoulder press. The explanation can include one or more parts (e.g., "Hold the dumbbells") in the list of parts. In this example, only the first part is displayed.

[0115] FIG. 3D illustrates an example interface for displaying a multi-part explanation for completing the exercise according to example embodiments of the present disclosure. In this example, the interface includes a general response 322 to the query. Below the general response 322 can be a list of parts to perform the task. Each part 334 in the multi-part list can be displayed in a distinct portion of the user interface.

[0116] Each portion of the list of parts can be expandable. In this example, the first part 336 in the list of parts has been expanded. For example, the user has clicked on a caret included in the expandible user interface element. In response, the portion of the user interface associated with the selected part expands, displaying additional information about that part. The additional data can include more information about the linked video. After the list of parts, the interface can also include an input field for asking follow-up questions. For example, if a user has a question about one of the parts or the general response, the user can enter their query in the input field 340. The query response system can then generate an appropriate response.

[0117] FIG. 4A illustrates an example interface for displaying a video captured by the camera of a user computing device accepting according to example embodiments of thepresent disclosure. A video display application can be used to capture video of a user performing a specific task. In this example, the user video display application can capture video of a scene as directed by the user of the user computing device. In this example, the scene is of a vehicle engine. The user interface 404 of the video display application can include a video display area 406 and a query input button 402 that is user-selectable.

[0118] FIG. 4B illustrates an example interface for receiving a query from a user according to example embodiments of the present disclosure. In this example, a user may, while displaying video captured by the camera of their user computing device 102, use a search input interface element (e.g., input element 406 in FIG. 4 A) to cause the user interface to display a query input field 416.

[0119] The user can input query text into the user input field 416. The query text can be input using a virtual keyboard displayed in the user interface. Other input methods can be used (e.g., voice input, wireless keyboard, motion input, and so on). The user computing device can accept the query text and capture at least a portion of the video being displayed while the query text is being input. In some examples, the user can indicate which portion of the displayed video is associated with the query text.

[0120] In some examples, the video display application can include focus indicators 418 in the user interface. The focus indicators 418 can be associated with the particular portion of the video that is of interest to the user. In some examples, the user can control the direction of the camera of the user computing device to ensure that the key portions of the scene that are associated with any query text are within the bounds of the focus indicators.

[0121] The query response system can use the query text and the associated video to generate an input prompt for the multimodal machine-learned model. The prompt can also include relevant supplemental information. For example, the query response system can analyze the video to determine a topic and provide information about that topic as part of the input to the machine-learned model. In this example, the query text reads, "How do I jumpstart my car?" and the video depicts a portion of a particular vehicle. The query response system can analyze the imagery to determine the vehicle's make, model, and year and provide that information (as well as any other information stored about that vehicle) to the model as contextual information.

[0122] In some examples, the query response system can determine the overall condition of the vehicle based on the imagery and the scene included in the video. For example, if a portion is damaged or a component is missing, the query response system can include that information in the input to the multimodal machine-learned system.

[0123] FIG. 4C illustrates an example interface for displaying an explanation for performing an exercise properly according to example embodiments of the present disclosure. In this example, the interface includes an initial response 422, along with information about the tools and / or components needed to perform the indicated action. The initial response 422 can be displayed in a portion of the user interface below the video display portion.

[0124] FIG. 4D illustrates an example interface for displaying a multi-part explanation for completing the exercise according to example embodiments of the present disclosure. In this example, the interface includes an initial response 422 to the query. The initial response 422 can include an indication of the subject of the video and / or scene and a list of tools and components needed to complete the task. Below the initial response 422 can be a list of parts to perform the task. Each part 434 in the multi-part list can be displayed in a distinct portion of the user interface.

[0125] Each portion of the list of parts can be expandable. In this example, a first part 436 in the list of parts has been expanded. For example, the user has clicked on a caret included in the expandible user interface element. In response, the portion of the user interface associated with the part expands, displaying additional information about that part. The additional data can include more information about the linked video. After the list of parts, the interface can also include an input field for asking follow-up questions. For example, if a user has a question about one of the parts or the general response, they enter their query in the input field 440 and the query response system can generate an appropriate response.

[0126] FIG. 4E illustrates an example interface for displaying a multi-part explanation for completing the exercise according to example embodiments of the present disclosure. In this example, the interface includes a list of parts to perform the task. Each part 434 in the multi-part list can be displayed in a distinct portion of the user interface.

[0127] In this example, each part can include an image or portion of a video that is extracted from the submitted video or displays the same scene as the submitted video. The image (or video) can include annotations displayed over the image. For example, part 1 indicates that a first portion of the scene is associated with the first part. The image (or video) displayed along with the first part can include visual annotations that help the user understand the portion of the scene associated with the part and / or how to perform the part.

[0128] In this example, the first part is "Locate the battery and open the battery lid." The associated image includes a white outline 442 indicating a specific component in the sceneassociated with the text of the first part (e.g., the battery or battery lid). In this way, the user can more easily understand the steps and any actions the user needs to take.

[0129] FIG. 5A depicts an example interface for application that presents visual data captured by a camera of the user computing device and analyzes the contents of the visual data to provide services to the user in accordance with example embodiments of the present disclosure.

[0130] In this example, a user computing device 500 includes a user interface 502 that displays visual data (images or video captured by the camera associated with the user computer device 500) to the user. In some examples, the application can determine that there is text within the image 504.

[0131] The interface can include a query request interface element 506 (e.g., a button in the interface) is selected by the user, the application can enable the user to provide query text and select at least a portion of the visual data for submission to a multimodal machine- learned model. The result of this model can be displayed in the interface as shown in FIG.5B.

[0132] FIG. 5B includes an example interface that can display a multi-part explanation to the user in response to the query request submitted by the user in accordance with example embodiments here. In this example, the interface can be updated to display the multi-part response 508. In some examples, each part of the multi-part response can be displayed in a distinct, collapsible, section. In some examples, the interface can also include a link 510 to return to the image display.

[0133] FIG. 6 represents an example system for providing multi-part response to queries that include visual data according to example embodiments of the present disclosure. The query response system 600 includes a reception system 602, an analysis system 604, an input generation system 606, a multimodal response model 608, a generation system 610, and the display system 612.

[0134] The reception system 602 can receive a request or query from a user. The request or query can include a text query, a visual data query, or an audio input query. In other examples, the request can be determined based on analysis of the content of visual data. For example, a visual data display application can display a video in the display of a user computing device. In some examples, that video can be captured previously and is now being displayed to a user. In other examples, the display video is a live representation of visual data currently being captured by a camera included in the user computing device.

[0135] In some examples, if an application is displaying visual data (e.g., a video or an image), the application can include an interface element that enables the user to input a query. Selection of this interface element (e.g., a button) can display a query input field and, once a query has been input, generate a request to the reception system 602 to provide it with a response to the query. The query can include the query text and visual data displaying a particular scene.

[0136] The analysis system 404 can analyze the query. In some examples, the analysis system 604 can analyze the query and the associated visual data. In some examples, the query is included in the visual data. For example, the video can include an audible question from the user depicted in the video. The analysis system can analyze the query and the visual data content to determine a topic or subject of the query. For example, if the visual data includes imagery of a vehicle’s engine, the analysis system 604 can determine, based on the visual data and the query text, the query is associated with vehicle repair.

[0137] The analysis system 404 can transmit the query, the visual data, and any subject information determined by the analysis system 404 based on the content of the visual data or the query text to the input generation system 606. The input generation system 406 can generate an input to the response model6408. For example, the input can be a prompt to a large language model that includes instructions about the type of response to be output by the response model 608.

[0138] In some examples, the input generation system 406 can access context data stored in a context data storage 234. The context data can include information about the user submitting the query (with the express permission of the user), information about one or more subjects associated with the query, information requesting a specific type of output, (e.g., based on the type of display in which the results will be displayed), and any information about requirements or restrictions for the results such as content restrictions.

[0139] The input generation system 606 can then generate a prompt that includes, but is not limited to, the query text, the visual data, any context data from the context data storage 234 and any information about previous interactions with the user (e.g., if this query is part of a multi-query conversation, the user’s previous inputs and outputs can be supplied as context as well). The input can be transmitted to the response model 608 (e.g., a multimodal machine-learned model) as input. In response, the response model 608 can generate a structured, multi-part response n as output. In some examples, the output generated can be text based, image based, or audio based.

[0140] In some examples, the output is a multi-part explanation that explains how to accomplish a particular task to respond to the query. Each part can be displayed in a distinct section of the user interface. In some examples, each section of the user interface can be collapsible. When collapsed, the section associated with each part can include a general summary of the part. When a section is expanded (e.g., a user clicks on an element in the user interface that causes a particular section to expand) a more detailed explanation of how to accomplish that part can be displayed. Each part can include visual data that depicts the same scene as the received visual data. For example, the visual data associated with a particular part can be an image from the submitted query video with overlaid annotations that provide a clearer understanding of how a particular task is to be performed.

[0141] In some examples, the output can also include a general summary of the high- level strategy of the multi-part solution, such that the user can understand the context of each part. The high-level strategy can be displayed above the series of parts in the user interface. In some examples, complicated parts may have a series of sub-parts within them. Each subpart can be displayed in a collapsible section within the parent part. In this way, additional details can be provided for each part as necessary with a series of nested explanations.

[0142] The output of the response model 608 can be transmitted to the generation system 610. The generation system 610 can receive the output of the response model 608 and generate, based on the output, information to be used to display the parts correctly in the user interface. For example, the output of the response model 608 can include formatting data for displaying the parts in the multi-part explanation. In other examples, the generation system 610 can generate the formatting data.

[0143] In some examples, the output of the response model 608 may not include the final visual data to be displayed with a particular part. Instead, that output can include text and instructions to generate or modify visual data as needed to provide appropriate explanations. The generation system 610 can use the instructions from the response model 608 to generate an appropriate visual response (e.g., an annotated image). For example, the output of the response model 608 may determine the particular portion of the input visual data to be displayed with a part (e.g., a frame within the video data or a portion of the video) and instructions on how to modify that portion. For example, the output can include an indication of a particular frame in a video from the input visual data and instructions to highlight a particular portion of the engine displayed in the frame. The generation system 610 can use instructions generated by the response model 608 to generate an appropriate visual data and include it in the part for which it is associated. In some examples, the generation system 610can access a machine-learned model that has been trained to generate visual data based on the output of the response model 608.

[0144] Once the generation system 610 has completed generating all needed portions of the multi-part response, the portions can be transmitted to the display system 612. The display system can cause the multi-part response to be displayed. For example, if the application is running on a user computer device, the response can be displayed in the appropriate portion of the user interface associated with the user computing device. However, if the query system 400 is running at a remote server system, the information for displaying the multi-part response can be transmitted by the display system 612 to the appropriate requesting device.

[0145] FIG. 7 depicts a block diagram of an example computing system 100 that performs a method for generating multi-part explanations that includes visual data according to example embodiments of the present disclosure. The system 100 includes a user computing system 102, a server computing system 130, and / or a third computing system 150 that are communicatively coupled over a network 180.

[0146] The user computing system 102 can include any type of computing device, such as, for example, a personal computing device (e.g., laptop or desktop), a mobile computing device (e.g., smartphone or tablet), a gaming console or controller, a wearable computing device, an embedded computing device, or any other type of computing device.

[0147] The user computing system 102 includes one or more processors 112 and a memory 114. The one or more processors 112 can be any suitable processing device (e.g., a processor core, a microprocessor, an ASIC, a FPGA, a controller, a microcontroller, etc.) and can be one processor or a plurality of processors that are operatively connected. The memory 114 can include one or more non-transitory computer-readable storage mediums, such as RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. The memory 114 can store data 116 and instructions 118 which are executed by the processor 112 to cause the user computing system 102 to perform operations.

[0148] In some implementations, the user computing system 102 can store or include one or more machine-learned models 120. For example, the machine-learned models 120 can be or can otherwise include various machine-learned models such as neural networks (e.g., deep neural networks) or other types of machine-learned models, including non-linear models and / or linear models. Neural networks can include feed-forward neural networks, recurrent neural networks (e.g., long short-term memory recurrent neural networks), convolutional neural networks or other forms of neural networks.

[0149] In some implementations, the one or more machine-learned models 120 can be received from the server computing system 130 over network 180, stored in the user computing device memory 114, and then used or otherwise implemented by the one or more processors 112. In some implementations, the user computing system 102 can implement multiple parallel instances of a single machine-learned model 120 (e.g., to perform parallel machine-learned model processing across multiple instances of input data and / or detected features).

[0150] More particularly, the one or more machine-learned models 120 may include one or more detection models, one or more classification models, one or more segmentation models, one or more augmentation models, one or more generative models, one or more natural language processing models, one or more optical character recognition models, and / or one or more other machine-learned models. The one or more machine-learned models 120 can include one or more transformer models. The one or more machine-learned models 120 may include one or more neural radiance field models, one or more diffusion models, and / or one or more autoregressive language models.

[0151] The one or more machine-learned models 120 may be utilized to detect one or more object features. The detected object features may be classified and / or embedded. The classification and / or the embedding may then be utilized to perform a search to determine one or more search results. Alternatively and / or additionally, the one or more detected features may be utilized to determine an indicator (e.g., a user interface element that indicates a detected feature) is to be provided to indicate a feature has been detected. The user may then select the indicator to cause a feature classification, embedding, and / or search to be performed. In some implementations, the classification, the embedding, and / or the searching can be performed before the indicator is selected.

[0152] In some implementations, the one or more machine-learned models 120 can process image data, text data, audio data, and / or latent encoding data to generate output data that can include image data, text data, audio data, and / or latent encoding data. The one or more machine-learned models 120 may perform optical character recognition, natural language processing, image classification, object classification, text classification, audio classification, context determination, action prediction, image correction, image augmentation, text augmentation, sentiment analysis, object detection, error detection, inpainting, video stabilization, audio correction, audio augmentation, and / or data segmentation (e.g., mask based segmentation).

[0153] Additionally, or alternatively, one or more machine-learned models 140 can be included in or otherwise stored and implemented by the server computing system 130 that communicates with the user computing system 102 according to a client-server relationship. For example, the machine-learned models 140 can be implemented by the server computing system 140 as a portion of a web service (e.g., a viewfinder service, a visual search service, an image processing service, an ambient computing service, and / or an overlay application service). Thus, one or more models 120 can be stored and implemented at the user computing system 102 and / or one or more models 140 can be stored and implemented at the server computing system 130.

[0154] The user computing system 102 can also include one or more user input component 122 that receives user input. For example, the user input component 122 can be a touch-sensitive component (e.g., a touch-sensitive display screen or a touch pad) that is sensitive to the touch of a user input object (e.g., a finger or a stylus). The touch-sensitive component can serve to implement a virtual keyboard. Other example user input components include a microphone, a traditional keyboard, or other means by which a user can provide user input.

[0155] In some implementations, the user computing system can store and / or provide one or more user interfaces 124, which may be associated with one or more applications. The one or more user interfaces 124 can be configured to receive inputs and / or provide data for display (e.g., image data, text data, audio data, one or more user interface elements, an augmented-reality experience, a virtual reality experience, and / or other data for display. The one or more user interfaces 124 may be associated with one or more other computing systems (e.g., server computing system 130 and / or third party computing system 150). The user interfaces 124 can include a viewfinder interface, a search interface, a generative model interface, a social media interface, and / or a media content gallery interface.

[0156] The user computing system 102 may include and / or receive data from one or more sensors 126. The one or more sensors 126 may be housed in a housing component that houses the one or more processors 112, the memory 114, and / or one or more hardware components, which may store, and / or cause to perform, one or more software packets. The one or more sensors 126 can include one or more image sensors (e.g., a camera), one or more lidar sensors, one or more audio sensors (e.g., a microphone), one or more inertial sensors (e.g., inertial measurement unit), one or more biological sensors (e.g., a heart rate sensor, a pulse sensor, a retinal sensor, and / or a fingerprint sensor), one or more infrared sensors, one or more location sensors (e.g., GPS), one or more touch sensors (e.g., a conductive touchsensor and / or a mechanical touch sensor), and / or one or more other sensors. The one or more sensors can be utilized to obtain data associated with a user’s environment (e.g., an image of a user’s environment, a recording of the environment, and / or the location of the user).

[0157] The user computing system 102 may include, and / or pe part of, a user computing device 104. The user computing device 104 may include a mobile computing device (e.g., a smartphone or tablet), a desktop computer, a laptop computer, a smart wearable, and / or a smart appliance. Additionally, and / or alternatively, the user computing system may obtain from, and / or generate data with, the one or more one or more user computing devices 104. For example, a camera of a smartphone may be utilized to capture image data descriptive of the environment, and / or an overlay application of the user computing device 104 can be utilized to track and / or process the data being provided to the user. Similarly, one or more sensors associated with a smart wearable may be utilized to obtain data about a user and / or about a user’s environment (e.g., image data can be obtained with a camera housed in a user’s smart glasses). Additionally, and / or alternatively, the data may be obtained and uploaded from other user devices that may be specialized for data obtainment or generation.

[0158] The server computing system 130 includes one or more processors 132 and a memory 134. The one or more processors 132 can be any suitable processing device (e.g., a processor core, a microprocessor, an ASIC, an FPGA, a controller, a microcontroller, etc.) and can be one processor or a plurality of processors that are operatively connected. The memory 134 can include one or more non-transitory computer-readable storage mediums, such as RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. The memory 134 can store data 136 and instructions 138 which are executed by the processor 132 to cause the server computing system 130 to perform operations.

[0159] In some implementations, the server computing system 130 includes or is otherwise implemented by one or more server computing devices. In instances in which the server computing system 130 includes plural server computing devices, such server computing devices can operate according to sequential computing architectures, parallel computing architectures, or some combination thereof.

[0160] As described above, the server computing system 130 can store or otherwise include one or more machine-learned models 140. For example, the models 140 can be or can otherwise include various machine-learned models. Example machine-learned models include neural networks or other multi-layer non-linear models. Example neural networksinclude feed forward neural networks, deep neural networks, recurrent neural networks, and convolutional neural networks. Example models 140 are discussed with reference to Figure 8.

[0161] Additionally and / or alternatively, the server computing system 130 can include and / or be communicatively connected with a search engine 142 that may be utilized to crawl one or more databases (and / or resources). The search engine 142 can process data from the user computing system 102, the server computing system 130, and / or the third-party computing system 150 to determine one or more search results associated with the input data. The search engine 142 may perform term based search, label based search, Boolean based searches, image search, embedding based search (e.g., nearest neighbor search), multimodal search, and / or one or more other search techniques.

[0162] The server computing system 130 may store and / or provide one or more user interfaces 144 for obtaining input data and / or providing output data to one or more users. The one or more user interfaces 144 can include one or more user interface elements, which may include input fields, navigation tools, content chips, selectable tiles, widgets, data display carousels, dynamic animation, informational pop-ups, image augmentations, text-to-speech, speech-to-text, augmented-reality, virtual-reality, feedback loops, and / or other interface elements.

[0163] The user computing system 102 and / or the server computing system 130 can train the models 120 and / or 140 via interaction with the third party computing system 150 that is communicatively coupled over the network 180. The third party computing system 150 can be separate from the server computing system 130 or can be a portion of the server computing system 130. Alternatively and / or additionally, the third party computing system 150 may be associated with one or more web resources, one or more web platforms, one or more other users, and / or one or more contexts.

[0164] The third party computing system 150 can include one or more processors 152 and a memory 154. The one or more processors 152 can be any suitable processing device (e.g., a processor core, a microprocessor, an ASIC, a FPGA, a controller, a microcontroller, etc.) and can be one processor or a plurality of processors that are operatively connected. The memory 154 can include one or more non-transitory computer-readable storage mediums, such as RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, etc., and combinations thereof. The memory 154 can store data 156 and instructions 158 which are executed by the processor 152 to cause the third party computing system 150 to perform operations. In some implementations, the third party computing system 150 includes or is otherwise implemented by one or more server computing devices.

[0165] The network 180 can be any type of communications network, such as a local area network (e.g., intranet), wide area network (e.g., Internet), or some combination thereof and can include any number of wired or wireless links. In general, communication over the network 180 can be carried via any type of wired and / or wireless connection, using a wide variety of communication protocols (e.g., TCP / IP, HTTP, SMTP, FTP), encodings or formats (e.g., HTML, XML), and / or protection schemes (e.g., VPN, secure HTTP, SSL).

[0166] The machine-learned models described in this specification may be used in a variety of tasks, applications, and / or use cases.

[0167] In some implementations, the input to the machine-learned model(s) of the present disclosure can be image data. The machine-learned model(s) can process the image data to generate an output. As an example, the machine-learned model(s) can process the image data to generate an image recognition output (e.g., a recognition of the image data, a latent embedding of the image data, an encoded representation of the image data, a hash of the image data, etc.). As another example, the machine-learned model(s) can process the image data to generate an image segmentation output. As another example, the machine- learned model(s) can process the image data to generate an image classification output. As another example, the machine-learned model(s) can process the image data to generate an image data modification output (e.g., an alteration of the image data, etc.). As another example, the machine-learned model(s) can process the image data to generate an encoded image data output (e.g., an encoded and / or compressed representation of the image data, etc.). As another example, the machine-learned model(s) can process the image data to generate an upscaled image data output. As another example, the machine-learned model(s) can process the image data to generate a prediction output.

[0168] In some implementations, the input to the machine-learned model(s) of the present disclosure can be text or natural language data. The machine-learned model(s) can process the text or natural language data to generate an output. As an example, the machine- learned model(s) can process the natural language data to generate a language encoding output. As another example, the machine-learned model(s) can process the text or natural language data to generate a latent text embedding output. As another example, the machine- learned model(s) can process the text or natural language data to generate a translation output. As another example, the machine-learned model(s) can process the text or natural language data to generate a classification output. As another example, the machine-learned model(s) can process the text or natural language data to generate a textual segmentation output. As another example, the machine-learned model(s) can process the text or naturallanguage data to generate a semantic intent output. As another example, the machine-learned model(s) can process the text or natural language data to generate an upscaled text or natural language output (e.g., text or natural language data that is higher quality than the input text or natural language, etc.). As another example, the machine-learned model(s) can process the text or natural language data to generate a prediction output.

[0169] In some implementations, the input to the machine-learned model(s) of the present disclosure can be speech data. The machine-learned model(s) can process the speech data to generate an output. As an example, the machine-learned model(s) can process the speech data to generate a speech recognition output. As another example, the machine- learned model(s) can process the speech data to generate a speech translation output. As another example, the machine-learned model(s) can process the speech data to generate a latent embedding output. As another example, the machine-learned model(s) can process the speech data to generate an encoded speech output (e.g., an encoded and / or compressed representation of the speech data, etc.). As another example, the machine-learned model(s) can process the speech data to generate an upscaled speech output (e.g., speech data that is higher quality than the input speech data, etc.). As another example, the machine-learned model(s) can process the speech data to generate a textual representation output (e.g., a textual representation of the input speech data, etc.). As another example, the machine- learned model(s) can process the speech data to generate a prediction output.

[0170] In some implementations, the input to the machine-learned model(s) of the present disclosure can be sensor data. The machine-learned model(s) can process the sensor data to generate an output. As an example, the machine-learned model(s) can process the sensor data to generate a recognition output. As another example, the machine-learned model(s) can process the sensor data to generate a prediction output. As another example, the machine-learned model(s) can process the sensor data to generate a classification output. As another example, the machine-learned model(s) can process the sensor data to generate a segmentation output. As another example, the machine-learned model(s) can process the sensor data to generate a segmentation output. As another example, the machine-learned model(s) can process the sensor data to generate a visualization output. As another example, the machine-learned model(s) can process the sensor data to generate a diagnostic output. As another example, the machine-learned model(s) can process the sensor data to generate a detection output.

[0171] In some cases, the input includes visual data and the task is a computer vision task. In some cases, the input includes pixel data for one or more images and the task is animage processing task. For example, the image processing task can be image classification, where the output is a set of scores, each score corresponding to a different object class and representing the likelihood that the one or more images depict an object belonging to the object class. The image processing task may be object detection, where the image processing output identifies one or more regions in the one or more images and, for each region, a likelihood that region depicts an object of interest. As another example, the image processing task can be image segmentation, where the image processing output defines, for each pixel in the one or more images, a respective likelihood for each category in a predetermined set of categories. For example, the set of categories can be foreground and background. As another example, the set of categories can be object classes. As another example, the image processing task can be depth estimation, where the image processing output defines, for each pixel in the one or more images, a respective depth value. As another example, the image processing task can be motion estimation, where the network input includes multiple images, and the image processing output defines, for each pixel of one of the input images, a motion of the scene depicted at the pixel between the images in the network input.

[0172] The user computing system may include a number of applications (e.g., applications 1 through N). Each application may include its own respective machine-learned library and machine-learned model(s). For example, each application can include a machine- learned model. Example applications include a text messaging application, an email application, a dictation application, a virtual keyboard application, a browser application, etc.

[0173] Each application can communicate with a number of other components of the computing device, such as, for example, one or more sensors, a context manager, a device state component, and / or additional components. In some implementations, each application can communicate with each device component using an API (e.g., a public API). In some implementations, the API used by each application is specific to that application.

[0174] The user computing system 102 can include a number of applications (e.g., applications 1 through N). Each application is in communication with a central intelligence layer. Example applications include a text messaging application, an email application, a dictation application, a virtual keyboard application, a browser application, etc. In some implementations, each application can communicate with the central intelligence layer (and model(s) stored therein) using an API (e.g., a common API across all applications).

[0175] The central intelligence layer can include a number of machine-learned models. For example, a respective machine-learned model (e.g., a model) can be provided for each application and managed by the central intelligence layer. In other implementations, two ormore applications can share a single machine-learned model. For example, in some implementations, the central intelligence layer can provide a single model (e.g., a single model) for all of the applications. In some implementations, the central intelligence layer is included within or otherwise implemented by an operating system of the computing system 100.

[0176] The central intelligence layer can communicate with a central device data layer. The central device data layer can be a centralized repository of data for the computing system 100. The central device data layer may communicate with a number of other components of the computing device, such as, for example, one or more sensors, a context manager, a device state component, and / or additional components. In some implementations, the central device data layer can communicate with each device component using an API (e.g., a private API).

[0177] FIG. 8 depicts a block diagram of an example computing system 50 that performs a process for generating multi-part responses to video queries according to example embodiments of the present disclosure. In particular, the example computing system 50 can include one or more computing devices 52 that can be utilized to obtain, and / or generate, one or more datasets that can be processed by a sensor processing system 60 and / or an output determination system 80 to feedback to a user that can provide information on features in the one or more obtained datasets. The one or more datasets can include image data, text data, audio data, multimodal data, latent encoding data, etc. The one or more datasets may be obtained via one or more sensors associated with the one or more computing devices 52 (e.g., one or more sensors in the computing device 52). Additionally and / or alternatively, the one or more datasets can be stored data and / or retrieved data (e.g., data retrieved from a web resource). For example, images, text, and / or other content items may be interacted with by a user. The interacted with content items can then be utilized to generate one or more determinations.

[0178] The one or more computing devices 52 can obtain, and / or generate, one or more datasets based on image capture, sensor tracking, data storage retrieval, content download (e.g., downloading an image or other content item via the internet from a web resource), and / or via one or more other techniques. The one or more datasets can be processed with a sensor processing system 60. The sensor processing system 60 may perform one or more processing techniques using one or more machine-learned models, one or more search engines, and / or one or more other processing techniques. The one or more processing techniques can be performed in any combination and / or individually. The one or more processing techniques can be performed in series and / or in parallel. In particular, the one ormore datasets can be processed with a context determination block 62, which may determine a context associated with one or more content items. The context determination block 62 may identify and / or process metadata, user profile data (e.g., preferences, user search history, user browsing history, user purchase history, and / or user input data), previous interaction data, global trend data, location data, time data, and / or other data to determine a particular context associated with the user. The context can be associated with an event, a determined trend, a particular action, a particular type of data, a particular environment, and / or another context associated with the user and / or the retrieved or obtained data.

[0179] The sensor processing system 60 may include an image preprocessing block 64. The image preprocessing block 64 may be utilized to adjust one or more values of an obtained and / or received image to prepare the image to be processed by one or more machine-learned models and / or one or more search engines 74. The image preprocessing block 64 may resize the image, adjust saturation values, adjust resolution, strip and / or add metadata, and / or perform one or more other operations.

[0180] In some implementations, the sensor processing system 60 can include one or more machine-learned models, which may include a detection model 66, a segmentation model 68, a classification model 70, an embedding model 72, and / or one or more other machine-learned models. For example, the sensor processing system 60 may include one or more detection models 66 that can be utilized to detect particular features in the processed dataset. In particular, one or more images can be processed with the one or more detection models 66 to generate one or more bounding boxes associated with detected features in the one or more images.

[0181] Additionally and / or alternatively, one or more segmentation models 68 can be utilized to segment one or more portions of the dataset from the one or more datasets. For example, the one or more segmentation models 68 may utilize one or more segmentation masks (e.g., one or more segmentation masks manually generated and / or generated based on the one or more bounding boxes) to segment a portion of an image, a portion of an audio file, and / or a portion of text. The segmentation may include isolating one or more detected objects and / or removing one or more detected objects from an image.

[0182] The one or more classification models 70 can be utilized to process image data, text data, audio data, latent encoding data, multimodal data, and / or other data to generate one or more classifications. The one or more classification models 70 can include one or more image classification models, one or more object classification models, one or more text classification models, one or more audio classification models, and / or one or more otherclassification models. The one or more classification models 70 can process data to determine one or more classifications.

[0183] In some implementations, data may be processed with one or more embedding models 72 to generate one or more embeddings. For example, one or more images can be processed with the one or more embedding models 72 to generate one or more image embeddings in an embedding space. The one or more image embeddings may be associated with one or more image features of the one or more images. In some implementations, the one or more embedding models 72 may be configured to process multimodal data to generate multimodal embeddings. The one or more embeddings can be utilized for classification, search, and / or learning embedding space distributions.

[0184] The sensor processing system 60 may include one or more search engines 74 that can be utilized to perform one or more searches. The one or more search engines 74 may crawl one or more databases (e.g., one or more local databases, one or more global databases, one or more private databases, one or more public databases, one or more specialized databases, and / or one or more general databases) to determine one or more search results. The one or more search engines 74 may perform feature matching, text based search, embedding based search (e.g., k-nearest neighbor search), metadata based search, multimodal search, web resource search, image search, text search, and / or application search.

[0185] Additionally and / or alternatively, the sensor processing system 60 may include one or more multimodal processing blocks 76, which can be utilized to aid in the processing of multimodal data. The one or more multimodal processing blocks 76 may include generating a multimodal query and / or a multimodal embedding to be processed by one or more machine-learned models and / or one or more search engines 74.

[0186] The output(s) of the sensor processing system 60 can then be processed with an output determination system 80 to determine one or more outputs to provide to a user. The output determination system 80 may include heuristic based determinations, machine-learned model based determinations, user selection based determinations, and / or context based determinations.

[0187] The output determination system 80 may determine how and / or where to provide the one or more search results in a search results interface 82. Additionally and / or alternatively, the output determination system 80 may determine how and / or where to provide the one or more machine-learned model outputs in a machine-learned model output interface 84. In some implementations, the one or more search results and / or the one or more machine-learned model outputs may be provided for display via one or more user interfaceelements. The one or more user interface elements may be overlayed over displayed data. For example, one or more detection indicators may be overlayed over detected objects in a viewfinder. The one or more user interface elements may be selectable to perform one or more additional searches and / or one or more additional machine-learned model processes. In some implementations, the user interface elements may be provided as specialized user interface elements for specific applications and / or may be provided uniformly across different applications. The one or more user interface elements can include pop-up displays, interface overlays, interface tiles and / or chips, carousel interfaces, audio feedback, animations, interactive widgets, and / or other user interface elements.

[0188] Additionally and / or alternatively, data associated with the output(s) of the sensor processing system 60 may be utilized to generate and / or provide an augmented-reality experience and / or a virtual -reality experience 86. For example, the one or more obtained datasets may be processed to generate one or more augmented-reality rendering assets and / or one or more virtual -reality rendering assets, which can then be utilized to provide an augmented-reality experience and / or a virtual-reality experience 86 to a user. The augmented- reality experience may render information associated with an environment into the respective environment. Alternatively and / or additionally, objects related to the processed dataset(s) may be rendered into the user environment and / or a virtual environment. Rendering dataset generation may include training one or more neural radiance field models to learn a three- dimensional representation for one or more objects.

[0189] In some implementations, one or more action prompts 88 may be determined based on the output(s) of the sensor processing system 60. For example, a search prompt, a purchase prompt, a generate prompt, a reservation prompt, a call prompt, a redirect prompt, and / or one or more other prompts may be determined to be associated with the output(s) of the sensor processing system 60. The one or more action prompts 88 may then be provided to the user via one or more selectable user interface elements. In response to a selection of the one or more selectable user interface elements, a respective action of the respective action prompt may be performed (e.g., a search may be performed, a purchase application programming interface may be utilized, and / or another application may be opened).

[0190] In some implementations, the one or more datasets and / or the output(s) of the sensor processing system 60 may be processed with one or more generative models 90 to generate a model-generated content item that can then be provided to a user. The generation may be prompted based on a user selection and / or may be automatically performed (e.g.,automatically performed based on one or more conditions, which may be associated with a threshold amount of search results not being identified).

[0191] The one or more generative models 90 can include language models (e.g., large language models and / or vision language models), image generation models (e.g., text-to- image generation models and / or image augmentation models), audio generation models, video generation models, graph generation models, and / or other data generation models (e.g., other content generation models). The one or more generative models 90 can include one or more transformer models, one or more convolutional neural networks, one or more recurrent neural networks, one or more feedforward neural networks, one or more generative adversarial networks, one or more self-attention models, one or more embedding models, one or more encoders, one or more decoders, and / or one or more other models. In some implementations, the one or more generative models 90 can include one or more autoregressive models (e.g., a machine-learned model trained to generate predictive values based on previous behavior data) and / or one or more diffusion models (e.g., a machine- learned model trained to generate predicted data based on generating and processing distribution data associated with the input data).

[0192] The one or more generative models 90 can be trained to process input data and generate model -generated content items, which may include a plurality of predicted words, pixels, signals, and / or other data. The model-generated content items may include novel content items that are not the same as any pre-existing work. The one or more generative models 90 can leverage learned representations, sequences, and / or probability distributions to generate the content items, which may include phrases, storylines, settings, objects, characters, beats, lyrics, and / or other aspects that are not included in pre-existing content items.

[0193] The one or more generative models 90 may include a vision language model.

[0194] The vision language model can be trained, tuned, and / or configured to process image data and / or text data to generate a natural language output. The vision language model may leverage a pre-trained large language model (e.g., a large autoregressive language model) with one or more encoders (e.g., one or more image encoders and / or one or more text encoders) to provide detailed natural language outputs that emulate natural language composed by a human.

[0195] The vision language model may be utilized for zero-shot image classification, few shot image classification, image captioning, multimodal query distillation, multimodal question and answering, and / or may be tuned and / or trained for a plurality of different tasks.The vision language model can perform visual question answering, image caption generation, feature detection (e.g., content monitoring (e.g., for inappropriate content)), object detection, scene recognition, and / or other tasks.

[0196] The vision language model may leverage a pre-trained language model that may then be tuned for multimodality. Training and / or tuning of the vision language model can include image-text matching, masked-language modeling, multimodal fusing with cross attention, contrastive learning, prefix language model training, and / or other training techniques. For example, the vision language model may be trained to process an image to generate predicted text that is similar to ground truth text data (e.g., a ground truth caption for the image). In some implementations, the vision language model may be trained to replace masked tokens of a natural language template with textual tokens descriptive of features depicted in an input image. Alternatively and / or additionally, the training, tuning, and / or model inference may include multi-layer concatenation of visual and textual embedding features. In some implementations, the vision language model may be trained and / or tuned via jointly learning image embedding and text embedding generation, which may include training and / or tuning a system to map embeddings to a joint feature embedding space that maps text features and image features into a shared embedding space. The joint training may include image-text pair parallel embedding and / or may include triplet training. In some implementations, the images may be utilized and / or processed as prefixes to the language model.

[0197] The output determination system 80 may process the one or more datasets and / or the output(s) of the sensor processing system 60 with a data augmentation block 92 to generate augmented data. For example, one or more images can be processed with the data augmentation block 92 to generate one or more augmented images. The data augmentation can include data correction, data cropping, the removal of one or more features, the addition of one or more features, a resolution adjustment, a lighting adjustment, a saturation adjustment, and / or other augmentation.

[0198] In some implementations, the one or more datasets and / or the output(s) of the sensor processing system 60 may be stored based on a data storage block 94 determination.

[0199] The output(s) of the output determination system 80 can then be provided to a user via one or more output components of the user computing device 52. For example, one or more user interface elements associated with the one or more outputs can be provided for display via a visual display of the user computing device 52.

[0200] The processes may be performed iteratively and / or continuously. One or more user inputs to the provided user interface elements may condition and / or affect successive processing loops.

[0201] FIG. 9 depicts an example flow diagram for a method of generating a multi-part response that include annotated video data according to example embodiments of the present disclosure. One or more portion(s) of the method can be implemented by one or more computing devices such as, for example, the computing devices described herein. Moreover, one or more portion(s) of the method can be implemented as an algorithm on the hardware components of the device(s) described herein. FIG. 9 depicts elements performed in a particular order for purposes of illustration and discussion. Those of ordinary skill in the art, using the disclosures provided herein, will understand that the elements of any of the methods discussed herein can be adapted, rearranged, expanded, omitted, combined, and / or modified in various ways without deviating from the scope of the present disclosure. The method can be implemented by one or more computing devices, such as one or more of the computing devices depicted in FIGS. 1 and 6.

[0202] In some examples, a query response system can, at 902, access visual data depicting a scene and a query associated with the visual data. In some examples, the visual data and query are received based on user input at a user computing system. In some examples, the query can include one or more of text data, image data, and audio data. The visual data can include at least one of a video or an image. In some examples, the video or image are captured by the camera integrated into the user computing device.

[0203] In some examples, the query response system can, at 904, generate input to a machine-learned model based on the visual data and the query. In some examples, the machine-learned model is a large language model. In some examples, the query response system can generate a machine-learned model prompt, wherein the prompt includes the query data, context information for the query data, and instructions to the machine-learned model. The contextual information can include user profile data describing a user current level of understanding.

[0204] In some examples, the query response system can, at 906, receive, as output from the machine-learned model, a multi-part response to the query, the multi-part response comprising a visual indication of the part to be performed with respect to the scene. The at least one part in the multi-part response can include a visual depiction of a process to be performed and text describing the process. In some examples, the visual depiction can comprise a visual indication overlaid on the visual data. In some examples, the visualdepiction can show an action to be performed or highlight a relevant portion of the scene. In some examples, the visual depiction visually represents the text of the part.

[0205] In some examples, the output of the machine-learned model includes formatting data for use in displaying the multi-part response. The formatting data can include markup data. The markup data can include, in markup language, directions describing how to display each part in the multi-part response as well as the layout of any content.

[0206] In some examples, the formatting data can cause each part in the multi-part response to be displayed in a distinct section of a user interface. Each distinct section of the user interface can be collapsible such that one or more parts in the multi-part response can be hidden. In some examples, the query response system can, at 908, provide the multi-part response for display to the user.

[0207] FIG. 10 depicts an example flow diagram for a method of generating multi -part responses to videos queries according to example embodiments of the present disclosure. One or more portion(s) of the method can be implemented by one or more computing devices such as, for example, the computing devices described herein. Moreover, one or more portion(s) of the method can be implemented as an algorithm on the hardware components of the device(s) described herein. FIG. 10 depicts elements performed in a particular order for purposes of illustration and discussion. Those of ordinary skill in the art, using the disclosures provided herein, will understand that the elements of any of the methods discussed herein can be adapted, rearranged, expanded, omitted, combined, and / or modified in various ways without deviating from the scope of the present disclosure. The method can be implemented by one or more computing devices, such as one or more of the computing devices depicted in FIGS. 1 and 6.

[0208] A computing system (e.g., user computing device 102 in FIG. 1) can include one or more processors, memory, and one or more communication systems. The one or more communication systems allow the computing system to transmit data to other computing systems via a communication network. The user computing device 102 (e.g., user computing device 102 in FIG. 1) can include other components that, together, enable the user computing device 102 (e.g., user computing device 102 in FIG. 1) to generate (or receive) a multi-part response to a query.

[0209] The computing system can, at 1002, receive a query from a user, wherein the query includes a video. In some examples, the query can be embedded in the audio of the video. For example, the user can, while recording video with a camera integrated into their smartphone (or other user computing device) audibly ask a question. In some examples, theuser computing device can, in real-time, analyze the audio of a video that is currently being captured. If the audio includes a query, the query response system can display a selectable element in the user interface (e.g., a popup button) that the user can select to receive a response to their question. In this way, the query response system can determine that a potential query is being asked automatically, without a user initially indicating that a query is being submitted. Such an automatic analysis would only be possible while executing a specific application (e.g., a video display and analysis application) and only with the explicit permission of the user. Doing so in a general context would be extremely expensive and unreliable.

[0210] In some examples, the query response system can, at 1004, generate input to a machine-learned model, the input including the video and a query request based on the query. In some examples, the input is a prompt for a multimodal large language model. The prompt can include the video, query text, contextual information about the query subject, contextual information about the user (if the user give permission), contextual information about previous queries form the user (including queries that are part of the same conversation as the current query), information about the s structure and format of the requested response, and other components of a prompt that are customized to the particular multimodal large language model.

[0211] In some examples, the query response system can, at 1004, receive as output from the machine-learned model, a multi-part response, wherein the multi-part response provides instructions associated with performing one or more actions and at least one part in the multi-part response includes text data and visual data. In some examples, the visual data can be videos or images extracted from the received video. In other examples, the visual data is not extracted from the received video.

[0212] In some examples, the query response system can, at 1008, provide the multi-part response for display to a user. The multi-part response can include a plurality of distinct collapsable sections. Each collapsable section can be associated with a part in the multi-part response. One or more parts can include a video. In some examples, the video can depict a user performing an action and the query request is associated with requesting instructions to improve the performance of the action.

[0213] The technology discussed herein makes reference to servers, databases, software applications, and other computer-based systems, as well as actions taken, and information sent to and from such systems. The inherent flexibility of computer-based systems allows for a great variety of possible configurations, combinations, and divisions of tasks andfunctionality between and among components. For instance, processes discussed herein can be implemented using a single device or component or multiple devices or components working in combination. Databases and applications can be implemented on a single system or distributed across multiple systems. Distributed components can operate sequentially or in parallel.

[0214] While the present subject matter has been described in detail with respect to various specific example embodiments thereof, each example is provided by way of explanation, not limitation of the disclosure. Those skilled in the art, upon attaining an understanding of the foregoing, can readily produce alterations to, variations of, and equivalents to such embodiments. Accordingly, the subject disclosure does not preclude inclusion of such modifications, variations and / or additions to the present subject matter as would be readily apparent to one of ordinary skill in the art. For instance, features illustrated or described as part of one embodiment can be used with another embodiment to yield a still further embodiment. Thus, it is intended that the present disclosure cover such alterations, variations, and equivalents.

Claims

WHAT IS CLAIMED IS:

1. A computer-implemented method comprising: accessing, by a computing system with one or more processors, visual data depicting a scene and a query associated with the visual data; generating, by the computing system, an input to a machine-learned model based on the visual data and the query; receiving, by the computing system as output from the machine-learned model, a multipart response to the query, the multi-part response comprising a visual indication of the part to be performed with respect to the scene; and providing, by the computing system, the multi-part response for display.

2. The computer-implemented method of claim 1, wherein the query includes one or more of: text data, image data, and audio data.

3. The computer-implemented method of claim 1, wherein the visual data comprises at least one of a video or an image.

4. The computer-implemented method of claim 1, wherein accessing, by a computing system with one or more processors, visual data depicting a scene and a query associated with the visual data comprises: receiving, by the computing system, the visual data and the query from a user computing device based on the input of a user.

5. The computer-implemented method of claim 1, wherein the machine-learned model is a large language model.

6. The computer-implemented method of claim 1, wherein at least one respective part in the multi-part response includes a visual depiction of a process to be performed and text describing the process.

7. The computer-implemented method of claim 6, wherein the visual depiction comprises a visual indication overlaid on the visual data.

8. The computer-implemented method of claim 7, wherein the visual indication highlights a portion of the visual data associated with the respective part.

9. The computer-implemented method of claim 7, wherein the visual indication depicts, with respect to the scene in the visual data, an action to be performed for the respective part.

10. The computer-implemented method of claim 7, wherein the visual indication is a visual depiction of the text of the respective part.

11. The computer-implemented method of claim 1, wherein the output of the machine- learned model includes formatting data for use in displaying the multi-part response.

12. The computer-implemented method of claim 11, wherein the formatting data includes markup data.

13. The computer-implemented method of claim 11, wherein the formatting data causes each part in the multi-part response to be displayed in a distinct section of a user interface.

14. The computer-implemented method of claim 13, wherein each distinct section of the user interface is collapsible such that one or more parts in the multi-part response can be hidden.

15. The computer-implemented method of claim 1, wherein generating, by the computing system, input to a machine-learned model based on the visual data and the query further comprises: generating a machine-learned model prompt, wherein the prompt includes the query data, context information for the query data, and instructions to the machine-learned model.

16. The computer-implemented method of claim 15, wherein the context information includes user profile data describing a user current level of understanding.

17. One or more non-transitory computer-readable media that collectively store instructions that, when executed by one or more computing devices, cause the one or more computing devices to perform operations, the operations comprising:accessing visual data depicting a scene and a query associated with the visual data; generating input to a machine-learned model based on the visual data and the query; receiving, as output from the machine-learned model, a multi-part response to the query, the multi-part response comprising a visual indication of the part to be performed with respect to the scene; and providing the multi-part response for display.

18. A computing system, the system comprising: one or more processors; and one or more non-transitory computer-readable media that collectively store instructions that, when executed by the one or more processors, cause the computing system to perform operations, the operations comprising: receiving a query from a user, wherein the query includes a video; generating input to a machine-learned model, the input including the video and a query request based on the query; receiving, as output from the machine-learned model, a multi-part response, wherein the multi-part response provides instructions associated with performing one or more actions and at least one part in the multi-part response includes text data and visual data; and providing the multi-part response for display to a user.

19. The computing system of claim 18, wherein the output includes, for each respective part of the multi-part response, a segment of the video associated with the respective part.

20. The computing system of claim 19, further comprises: analyzing the video to determine a plurality of distinct segments within the video; determining one or more features of each segment in the plurality of segments; and for a respective part in the multi-part response: associating the respective part of the multi-part response with a respective segment in the plurality of segments based on the one or more features of the respective segment.

21. The computing system of claim 20, wherein the determining one or more features of each segment in the plurality of segments further comprises: determining a topic associated with each segment of the plurality of distinct segments.

22. The computing system of claim 18, wherein the output includes a video with multiple segments and one or more parts of the multi-part response are associated with a particular segment in the video with multiple segments.

23. The computing system of claim 18, wherein the video depicts a user performing an action and the query request is associated with requesting instructions to improve the performance of the action.

Images