An interactive trial play advertisement automatic generation method and system based on structured analysis

Abstract

The application provides an interactive trial play advertisement automatic generation method and system based on structured analysis, a computing device and a computer readable storage medium. The method adopts a multi-modal joint recognition technology, fuses visual information and time sequence information and text set information to generate a time axis label with semantics, and accurately locates key interaction nodes in a video. Subsequently, an interactive graph intermediate representation is innovatively introduced, an event sequence is automatically arranged into game logic which can be compiled across terminals through an intelligent template matching algorithm, and decoupling of logic definition and platform implementation is realized. The required interactive material is obtained through video reverse reconstruction. The application not only shortens the delivery cycle from "day / week" level to "minute / hour" level, greatly reduces the cost and realizes large-scale production, but also ensures the coherence of the interactive logic and the video content, and significantly improves the playability and advertisement conversion effect.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a method and system for automatically generating interactive playable advertisements based on structured analysis, a computing device, and a computer-readable storage medium. Background Technology

[0002] Playable ads, as a form of advertising that allows users to engage in lightweight interactive experiences before downloading an app, have become a key tool for mobile app promotion and user growth, aiming to significantly improve ad click-through rates and conversion rates. However, current mainstream methods for creating playable ads still rely heavily on manual or semi-automated processes. These typically involve tedious steps such as video script breakdown, art asset slicing, H5 or Unity interactive programming, and manual testing. This traditional method has significant technical drawbacks: first, it heavily relies on manual labor and cross-functional collaboration, requiring repeated communication and iteration among editors, artists, and developers, resulting in long delivery cycles and high labor costs; second, it lacks reusability and scalability, as different video themes and platform specifications vary greatly, leading to repetitive work and hindering large-scale, low-cost mass production; third, the interaction and storyline are disconnected, lacking a systematic semantic mapping from video content to interactive behavior, resulting in simple gameplay but limited improvement in ad effectiveness. Therefore, there is an urgent need for an efficient method for generating playable ads that can automatically convert ordinary video footage into high-quality playable ads. Summary of the Invention

[0003] In view of this, embodiments of this application provide a method and apparatus for automatically generating interactive playable advertisements based on structured analysis, a computing device and a computer-readable storage medium, to address the technical deficiencies existing in the prior art.

[0004] According to a first aspect of the embodiments of this application, a method for automatically generating interactive playable advertisements based on structured analysis is provided, including:

[0005] The input video material is preprocessed to obtain a video frame sequence and a set of timing information text.

[0006] Visual analysis is performed on the video frame sequence to obtain structured scene elements, and then the structured scene elements are fused with the temporal information text set to obtain semantic timeline labels;

[0007] A game logic framework instance is generated based on the semantic timeline tags, and an interaction diagram is generated based on the structured scene elements and the game logic framework instance.

[0008] Reconstruct interactive resource materials based on the structured scene elements described in the video;

[0009] The interactive graph and the interactive resource materials are compiled and processed to generate an interactive trial advertisement product package.

[0010] According to a second aspect of the embodiments of this application, an automatic interactive playable advertisement generation system based on structured analysis is provided, comprising:

[0011] The preprocessing unit is used to preprocess the input video material to obtain a video frame sequence and a set of timing information text.

[0012] A visual analysis unit is used to perform visual analysis on the video frame sequence to obtain structured scene elements;

[0013] The fusion unit is used to fuse the structured scene elements with the temporal information text set to obtain semantic timeline labels;

[0014] The generation unit is used to generate a game logic framework instance based on the semantic timeline labels, and to generate an interaction diagram based on the structured scene elements and the game logic framework instance.

[0015] A reconstruction unit is used to reconstruct interactive resource materials based on the structured scene elements in the video.

[0016] The compilation unit is used to compile and process the interactive graph and the interactive resource materials to generate an interactive trial advertisement product package.

[0017] According to a third aspect of the embodiments of this application, a computing device is provided, including a memory, a processor, and computer instructions stored in the memory and executable on the processor, wherein the processor executes the instructions to implement the steps of the method for automatically generating interactive playable advertisements based on structured analysis.

[0018] According to a fourth aspect of the embodiments of this application, a computer-readable storage medium is provided that stores computer instructions, which, when executed by a processor, implement the steps of the method for automatically generating interactive playable advertisements based on structured analysis.

[0019] The method provided in the embodiments of this application constructs an end-to-end intelligent pipeline from video parsing to product output. First, the input video material is preprocessed to obtain a video frame sequence and a temporal information text set. Structured scene elements are extracted from the video frame sequence. Then, multimodal joint recognition technology is used to fuse visual and temporal information text set information to generate semantically meaningful timeline labels, accurately locating key interactive nodes in the video. Subsequently, an innovative interactive graph intermediate representation is introduced. Through an intelligent template matching algorithm, the event sequence is automatically arranged into cross-platform compilable game logic, achieving decoupling between logic definition and platform implementation. For the required interactive materials, techniques such as instance segmentation and trajectory fitting are used to reconstruct interactive resources and physical attributes from the video. This invention not only shortens the delivery cycle from "days / weeks" to "minutes / hours," significantly reducing costs and enabling large-scale production, but also ensures the coherence of interactive logic and video content, significantly improving playability and advertising conversion effects. Attached Figure Description

[0020] Figure 1 This is a structural block diagram of the computing device provided in the embodiments of this application;

[0021] Figure 2 This is a flowchart illustrating an automatic generation method for interactive playable advertisements based on structured analysis provided in an embodiment of this application.

[0022] Figure 3 This is a schematic diagram of an interactive playable advertisement automatic generation system based on structured analysis provided in an embodiment of this application. Detailed Implementation

[0023] Many specific details are set forth in the following description to provide a full understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this application; therefore, this application is not limited to the specific embodiments disclosed below.

[0024] The terminology used in one or more embodiments of this application is for the purpose of describing particular embodiments only and is not intended to limit the scope of one or more embodiments of this application. The singular forms “a,” “the,” and “the” used in one or more embodiments of this application and in the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” used in one or more embodiments of this application refers to and includes any or all possible combinations of one or more associated listed items.

[0025] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this application, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first may also be referred to as second without departing from the scope of one or more embodiments of this application, and similarly, second may also be referred to as first. Depending on the context, the word "if" as used herein may be interpreted as "in response to a determination".

[0026] This application provides a method and system for automatically generating interactive playable advertisements based on structured analysis, as well as a computing device and a computer-readable storage medium, which will be described in detail in the following embodiments.

[0027] Figure 1 A structural block diagram of a computing device 100 according to an embodiment of this application is shown. The components of the computing device 100 include, but are not limited to, a memory 110 and a processor 120. The processor 120 is connected to the memory 110 via a bus 130, and a database 150 is used to store data.

[0028] The computing device 100 also includes an access device 140, which enables the computing device 100 to communicate via one or more networks 160. Examples of these networks include a Public Switched Telephone Network (PSTN), a Local Area Network (LAN), a Wide Area Network (WAN), a Personal Area Network (PAN), or a combination of communication networks such as the Internet. The access device 140 may include one or more of any type of wired or wireless network interface (e.g., a Network Interface Card (NIC)), such as an IEEE 802.11 Wireless Local Area Network (WLAN) interface, a Wi-MAX interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, a Near Field Communication (NFC) interface, and so on.

[0029] In one embodiment of this application, the aforementioned components of the computing device 100 and Figure 1 Other components, not shown, can also be connected to each other, for example, via a bus. It should be understood that... Figure 1 The block diagram of the computing device shown is for illustrative purposes only and is not intended to limit the scope of this application. Those skilled in the art can add or replace other components as needed.

[0030] The computing device 100 can be any type of stationary or mobile computing device, including mobile computers or mobile computing devices (e.g., tablet computers, personal digital assistants, laptop computers, notebook computers, netbooks, etc.), mobile phones (e.g., smartphones), wearable computing devices (e.g., smartwatches, smart glasses, etc.) or other types of mobile devices, or stationary computing devices such as desktop computers or PCs.

[0031] In the embodiments of this application, Figure 2 A flowchart of an automatic interactive playable ad generation method based on structured analysis provided in this application is shown, including steps 202 to 210.

[0032] Step 202: Preprocess the input video material to obtain a video frame sequence and a set of timing information text.

[0033] In this step of the embodiments of this application, input game promotional video material (e.g., digital multimedia files in MP4, AVI, or MOV format) is received. This video material can be either actual gameplay footage, i.e., actual gameplay captured using screen recording technology; or it can be a CG promotional video, i.e., a game trailer video that has been rendered and post-composited.

[0034] Furthermore, the video footage is parsed, breaking it down into video streams, audio streams, and metadata. Specifically, the data in the video footage is read and parsed, and demultiplexing technology is used to separate the interleaved data packets based on track identifiers. This extracts the video frame sequence, the independent audio stream, and metadata containing information such as duration, resolution, and encoding format, providing standardized input for subsequent decoding or analysis steps.

[0035] Preferably, in order to eliminate the differences in encoding parameters between video materials from different sources, reduce the computational load of subsequent models and improve processing efficiency, a standardization preprocessing is performed before parsing the video materials, and the frame rate of the video is resampled to a preset standard frame rate (e.g., uniformly adjusted to 30fps), which reduces data redundancy while ensuring smooth motion.

[0036] Furthermore, text information in the video frame sequence is extracted using OCR (Optical Character Recognition) technology, and speech-text information is extracted using ASR (Automatic Speech Recognition) technology, in preparation for subsequent multimodal analysis.

[0037] Specifically, OCR technology is used to scan images in a video frame sequence, such as the "Click to Start" button, score display, subtitles, and special effects text on the UI interface in a game recording video. For example, a text detection algorithm is used to locate the region of interest (ROI), and then a text recognition algorithm is used to convert the pixel information into a computer-readable text string and record its timestamp (e.g., the text "Click to Start" appears from 00:00:01 to 00:00:03). Acoustic Response Speech (ASR) is used to extract acoustic features from the audio stream, identify valid human voice segments, and use acoustic and language models to convert the speech waveform into text commands, such as converting "swipe left and right to avoid obstacles" into text, and simultaneously recording the start and end times of the speech on the timeline.

[0038] In this step, the original unstructured video file is parsed and converted into a preprocessed dataset containing the following:

[0039] The data includes a standardized video frame sequence, separated audio stream data, and a time-series information text set containing time-stamped UI text obtained from the video frame sequence and time-stamped speech text obtained from the audio stream.

[0040] Step 204: Perform visual analysis on the video frame sequence to obtain structured scene elements, and then fuse the structured scene elements with the temporal information text set to obtain semantic timeline labels.

[0041] In this embodiment, the video frame sequence obtained in step 202 is subjected to deep analysis to obtain structured scene elements in the video. These structured scene elements include semantic segments, interactive objects, and key events / actions of the interactive objects. Specifically, the identification of structured scene elements in the video frame sequence is achieved through the following steps:

[0042] Step 2042: Detect the shot boundaries of the video frame sequence and segment the video into different semantic segments.

[0043] In this process, algorithms, such as histogram difference analysis or convolutional neural network algorithms, are used to calculate the pixel differences between adjacent frames frame by frame. When the difference value exceeds a set threshold, it is determined to be a shot transition point. For example, when analyzing the video timeline, it was found that the content of the scene changed drastically between 1.9 seconds and 2.0 seconds (from the "main menu" screen to the "game scene" screen). Therefore, a "segmentation point" was set at 2.0 seconds, dividing the video into "Segment A: Main Menu (0-2 seconds)" and "Segment B: Game Process (2-6 seconds)".

[0044] Furthermore, semantic recognition is performed on the image content within each time slice, including but not limited to:

[0045] If a large number of UI controls (such as "Start" and "Exit" buttons) are detected without any game entity interaction, then the segment is marked as a UI state (such as Intro or Pause).

[0046] If the game's main character entity is detected and accompanied by continuous environmental movement or user interaction, it is marked as a game state (such as Level or Tutorial).

[0047] If score statistics or win / loss notification text are detected, the status is marked as settlement (e.g., Victory, Defeat).

[0048] Step 2044: Identify interactive objects in the video frame sequence.

[0049] In this step, an object detection and instance segmentation model is used to extract features from each frame of the image in different semantic segments, identifying and delineating interactive objects in the scene, such as the main character, buttons, obstacles, and rewards. The object detection and instance segmentation model can be built based on a deep convolutional neural network or a visual Transformer architecture, which will not be elaborated here.

[0050] Step 2046: Identify key events / actions of objects in the video frame sequence.

[0051] In this step, the movement trajectory and changes of interactive objects in consecutive frames are analyzed to identify key actions or events in the video frames, such as "jumping," "shooting," and "eliminating." For example, when analyzing a consecutive frame sequence from the 3rd second to the 4.5th second, the Y-axis coordinate of the "protagonist" object is captured as exhibiting a parabolic trajectory of "rapid ascent followed by slow descent." Simultaneously, the posture analysis model detects that the character's legs are in a tucked-in posture. Combining visual features, the model determines that the event occurring within this spatiotemporal window is "jumping," and at this point, structured data is generated: {Time range: 00:03:00 - 00:04:50, Subject: Protagonist, Action label: Jump, Trajectory: [(x1,y1), (x2,y2)...]}. Those skilled in the art should know that the above process can be implemented using 3D convolutional neural networks (such as C3D) or optical flow methods, which will not be elaborated further here.

[0052] Furthermore, in order to eliminate the ambiguity that may arise from a single data source (only visuals or only audio), precise, semantic timeline labels are constructed by fusing visual information and text sets with temporal information. These semantic timeline labels will serve as the real data source for generating game logic in the future.

[0053] Specifically, the text collection containing time-series information includes:

[0054] The text and its position on the UI screen obtained from video frame sequence recognition, such as: {content: "Click to start", time: 1.0s-3.0s}, {content: "Swipe right to avoid", time: 4.10s - 5.00s}, and the speech-to-text obtained from audio stream recognition, such as: {content: "Swipe to avoid", time: 4.6s}.

[0055] Furthermore, multimodal semantic fusion is performed on the key events / actions of the aforementioned objects and the text set with temporal information. Multimodal semantic fusion is a core concept in the field of artificial intelligence. It combines computer vision (seeing), speech recognition (hearing), and text recognition (reading). By analyzing the correlation between them, it derives accurate semantic meanings that cannot be obtained from a single modality, enabling machines to understand the world in a comprehensive way.

[0056] In a specific illustrative scenario, at the 4.5-second mark of the game video, an obstacle appears on the screen, and the player swipes right to avoid it. The data for each modality includes:

[0057] The visual modalities derived from video structured analysis are:

[0058] T = 03.00s: An obstacle has been detected entering the screen.

[0059] T = 04.50s: The detected subject is the main character, action tag: Jump.

[0060] The text modality derived from OCR recognition is:

[0061] T = 04.10s - 05.00s: The screen text "Swipe right to dodge" was detected, and the coordinates were located at the bottom of the screen.

[0062] The audio modalities derived from ASR recognition and sound effect detection are:

[0063] T = 04.40s: The voice content was recognized as "Get out of the way quickly".

[0064] T = 04.50s: Sound effect feature Whoosh_Sound was detected.

[0065] Then, the aforementioned visual modality, OCR-recognized text modality, and audio modality are subjected to multimodal semantic fusion. First, the data are aligned on the timeline, and cross-validation is performed:

[0066] Time alignment: Key information from all modalities was found to be highly concentrated within the 4.00s - 4.50s time window;

[0067] Semantic association: The visual jumping action of the protagonist is consistent with the meaning of "swipe right" in the text; and the audio "quickly dodge" and the visual obstacle form a causal logic.

[0068] After multimodal data fusion, semantic timeline labels are generated to mark various interactive atomic events that occur in the video, such as:

[0069] Event ID: E_005

[0070] Semantic tags: interaction_swipe right

[0071] Trigger time: 04.50s

[0072] Confidence level: 99% (high)

[0073] Description: The user executed the "avoid obstacle" command and swiped to the right.

[0074] Those skilled in the art should understand that the above-mentioned generated interactive atomic events are examples rather than exhaustive. In specific implementation processes, the embodiments of this application can generate multiple semantic timeline labels containing various interactive atomic events.

[0075] Step 206: Generate a game logic framework instance based on the semantic timeline labels, and generate an interaction diagram based on the structured scene elements and the game logic framework instance.

[0076] In this embodiment, a gameplay template library is pre-built and maintained, which stores various interaction logic templates corresponding to different game types. Each interaction logic template defines the following two parts:

[0077] State machine skeleton: Defines the logical topology and flow rules for this gameplay type. For example, for the "parkour" template, its state machine skeleton defines the basic state flow path as: idle -> running -> dodging obstacles / collecting rewards -> failure / settlement.

[0078] Set of configurable parameters: These parameters define variables that affect the game experience and numerical balance. Configurable parameters include at least: game duration, protagonist movement speed, obstacle generation density, target score, input latency, and physics engine parameters (such as gravity).

[0079] Specifically, this step receives the semantic timeline labels output from step 204 and extracts feature vectors from them to characterize the interaction patterns. These feature vectors contain at least:

[0080] Event type distribution: Statistics on the frequency and percentage of various interactive atomic events (such as click, swipe, long press) in the timeline;

[0081] Temporal rhythm characteristics: Calculate the time interval between adjacent interactive events to determine whether the operation rhythm is a high-frequency continuous click or an intermittent operation;

[0082] Spatial movement characteristics: Based on the visual analysis results, determine the main movement trajectory of the interactive object in the screen coordinate system (whether it is mainly vertical jumping or mainly horizontal planar movement).

[0083] Furthermore, the similarity between the extracted feature vector and the preset feature vectors of each template in the gameplay template library is calculated. In one embodiment, cosine similarity or Euclidean distance can be used as a metric, and the template with the highest similarity exceeding a preset threshold is selected as the target template; if the similarity does not exceed the threshold, the video gameplay is determined to be a new type, and a general basic template can be called or a manual intervention can be prompted to define a new template.

[0084] Then, after determining the target template, the set of configurable parameters in the target template is instantiated and populated. For example, the game duration in the template parameters can be directly set according to the total duration of the video, or the number of obstacles or target objects appearing per unit time in the video can be counted to calculate and set the obstacle generation density or target score parameters; or the protagonist's movement speed parameters can be calculated and set in reverse based on the background object movement pixel speed obtained from video analysis.

[0085] Furthermore, the instantiated template object containing specific state transition logic and parameter values ​​is deeply integrated with the semantic timeline tags of the video to generate a platform-independent executable intermediate representation, namely the interaction graph. This graph fully defines the user interaction, internal state transition, and feedback mechanisms within Playable, including:

[0086] S2062: Node construction and time mapping.

[0087] Based on the state machine skeleton in the instantiated template and the structured scene elements of the video obtained above, nodes of the interaction graph are constructed. Nodes represent various logical stages in the Playable's execution process.

[0088] Specifically, based on the semantic features of semantic segments obtained from shot boundary detection, a corresponding state node is generated for each segment. The attributes of the node include, but are not limited to: node ID, node type (such as Loading, Intro, Tutorial, Gameplay, Settlement), and duration. For example, if the video timeline shows "Tutorial Guidance" from 0s to 3s and "Core Gameplay" from 3s to 20s, the system generates corresponding Tutorial nodes (with a duration of 3 seconds) and Gameplay nodes.

[0089] S2064: Edge definition and triggering conditions are bound together.

[0090] Based on the flow rules in the instantiated template and the structured scene elements of the video obtained above, edges connecting the various nodes are constructed. Each edge contains a source node, a target node, and one or more trigger conditions. The generation logic of the trigger conditions is to convert the interactive atomic events in the video into logical judgment expressions. For example:

[0091] If a "click" event is detected in the video at T=5.0s, an edge is generated from the current state to the next state, and the trigger condition is bound to Input_Tap_Detected (click input detected).

[0092] If the node is set to automatically redirect after a timeout, the trigger condition is Time_Exceeds(3000ms).

[0093] If a "collision obstacle" event is detected in the video, an edge is generated from the Playing state to the Failed state, with the binding condition being Collision_Event(Type=Obstacle).

[0094] S2066. Motion choreography and parameter injection

[0095] This step defines specific actions on nodes and edges. Actions specify the specific operations that the system should perform at a particular time, including but not limited to rendering updates, physics calculations, data statistics, and sound effect playback.

[0096] Actions serve as the bridge connecting semantic video with the final runtime resources. The system maps visual / auditory events in the video to specific function calls. For example:

[0097] When entering a State_Jump node or when the jump trigger condition is met, an Action_PlayAnimation is generated, and the corresponding resource references (such as "main character jump sequence frames") and animation duration parameters are injected.

[0098] When an "explosion" visual feature appears in the video, an Action_SpawnEffect is generated under the corresponding node, specifying the name and coordinates of the effect's prefab resource.

[0099] When the floating text "+10" appears in the video, the system generates Action_UpdateScore and sets the parameter Delta=10.

[0100] Based on the audio / background sound track of the video, an Action_PlaySound is generated at a specific node and bound to the corresponding audio resource ID.

[0101] S2068. Serialization and Output of Interaction Graphs

[0102] After completing the above construction, the interaction graph, containing nodes, edges, actions, and their complete dependencies, is serialized into a structured data format (such as JSON, XML, or binary format). The generated interaction graph does not depend on a specific operating system or game engine, but serves as a standardized logical description for subsequent compiler modules to parse and translate into specific runtime code, achieving an automated conversion from video content to runnable logic.

[0103] Step 208: Reconstruct interactive resource materials based on the structured scene elements in the video.

[0104] In this step of the embodiment of this application, the resources and physical attributes required to construct Playable are extracted and generated from the original video footage. Since the original video is only a pixel sequence and lacks interactivity, it is necessary to separate, enhance, and physically reconstruct key objects in the video so that they can be invoked and controlled by the runtime engine.

[0105] Specifically, interactive objects, such as game characters and props, within the identified structured scene elements are extracted at the pixel level using deep learning algorithms. A reasonable process includes: using an instance segmentation network to accurately extract the contour mask of the foreground object, achieving separation from the complex background; then applying an intelligent matting algorithm to finely optimize the edges, eliminating jagged edges to generate a high-quality mask; finally, extracting the processed pixels and adding alpha channels, saving them as an image sequence in time series, thus forming independent image resource materials.

[0106] Furthermore, motion trajectory estimation and fitting are performed on the generated image resources. A reasonable process includes: tracking the positional changes of the center point or key feature points of interactive objects in consecutive video frames; smoothing and curve fitting the acquired discrete coordinate point sequence; and extracting keyframes based on motion curvature or time intervals to generate motion description data. For example, if the protagonist in the video performs a parabolic jump, the mathematical equation of the parabola is fitted and converted into motion tweening data or keyframe animation data in the game engine, so that the reconstructed material can reproduce the same motion path.

[0107] Preferably, in order to achieve physical interaction and collision detection, a physical bounding box is automatically generated as a collider based on the visual shape of the object. The generated collider data is then bound to the corresponding image resource ID and the physical type of the collider is marked, so as to accurately respond to interactive events such as clicks, touches or obstructions in the runtime engine.

[0108] In this step, interactive objects in the structured scene elements are extracted and dynamically analyzed, thereby transforming pixels in the non-interactive video into a complete set of materials containing appearance, motion, and physical attributes, providing the necessary data support for subsequent compilation and packaging.

[0109] Step 210: Compile and process the interactive graph and interactive resource materials to generate an interactive trial ad product package.

[0110] In this step, the abstract logic generated in the previous steps is integrated with the static resources, and a Playable ad package that can run independently is generated through compilation.

[0111] Specifically, the process involves reading the interaction graph and interactive resource materials generated in the preceding steps. A compiler is then used to traverse and parse the interaction graph, translating node, edge, and action definitions into executable script instructions or bytecode for the target platform. For example, state transition logic is generated as JavaScript code or C# script, and physical parameters are injected into the physics engine's configuration interface.

[0112] Furthermore, the compiler establishes a mapping index between logical instructions and interactive resource assets. For example, in the generated code, when the instruction to "play jump animation" is executed, the reconstructed resource data is accurately invoked through the index pointer.

[0113] Then, a pre-built, lightweight, and portable runtime engine is embedded into the target artifact package. This runtime engine is the core environment for Playable's execution, responsible for interpreting and executing the compiled logic instructions and providing basic system capabilities. Finally, the compiler generates a Playable artifact package that can be directly deployed. In one embodiment, this artifact package is in standard HTML5 format (containing HTML, CSS, JavaScript, and resource files), and can be directly embedded into a web browser or a mobile app's WebView to run.

[0114] Preferably, before the final output, a simulation robot automatically executes a comprehensive test script based on the interaction graph to simulate real user operations and verify the integrity and robustness of the interaction process. By automatically detecting key indicators such as first-frame latency, interaction success rate, and stuttering, a quality assurance closed loop is constructed; if the product does not meet the standards, the system will automatically trigger a backflow mechanism to instruct upstream modules to optimize and adjust.

[0115] Preferably, this application embodiment also provides a WYSIWYG visual editor, supporting manual fine-tuning of the automatically generated output, such as replacing materials or adjusting difficulty parameters. This process introduces a weakly supervised correction mechanism, converting all manual revisions into feedback samples and recording them for continuous learning and iterative optimization of the model, thereby achieving a closed-loop improvement in the quality of the generated output.

[0116] In the above embodiments of this application, to address key technical issues in existing Playable ad production processes, such as heavy reliance on manual labor, long production cycles, insufficient reusability, and the disconnect between interactive logic and video content, this application constructs an end-to-end intelligent pipeline from video parsing to Playable ad output. First, the input video material is preprocessed to obtain a video frame sequence and a temporal information text set. Structured scene elements are extracted from the video frame sequence, and then multimodal joint recognition technology is used to fuse visual and temporal information text set information to generate semantically meaningful timeline labels, accurately locating key interactive nodes in the video. Subsequently, an innovative interactive graph intermediate representation is introduced. Through an intelligent template matching algorithm, the event sequence is automatically arranged into cross-platform compilable game logic, achieving decoupling between logic definition and platform implementation. For the required interactive materials, instance segmentation, trajectory fitting, and collision inference technologies are used to reconstruct interactive resources and physical attributes from the video.

[0117] Furthermore, an automated simulation testing mechanism has been incorporated, using robots to simulate real interactions to verify logical robustness and performance metrics, thus constructing a quality assurance closed loop. Combined with weakly supervised feedback from a human-machine collaborative editor, the model's capabilities are continuously optimized. This invention not only shortens the delivery cycle from days / weeks to minutes / hours, significantly reducing costs and enabling large-scale production, but also ensures the consistency of interactive logic and video content, significantly improving playability and advertising conversion rates.

[0118] Corresponding to the above embodiment of an automatic interactive playable ad generation method based on structured analysis, this application also provides an embodiment of an automatic interactive playable ad generation system based on structured analysis, such as... Figure 3 As shown, the system includes:

[0119] The preprocessing unit is used to preprocess the input video material to obtain a video frame sequence and a set of timing information text.

[0120] A visual analysis unit is used to perform visual analysis on the video frame sequence to obtain structured scene elements;

[0121] The fusion unit is used to fuse the structured scene elements with the temporal information text set to obtain semantic timeline labels;

[0122] The generation unit is used to generate a game logic framework instance based on the semantic timeline labels, and to generate an interaction diagram based on the structured scene elements and the game logic framework instance.

[0123] A reconstruction unit is used to reconstruct interactive resource materials based on the structured scene elements in the video.

[0124] The compilation unit is used to compile and process the interactive graph and the interactive resource materials to generate an interactive trial advertisement product package.

[0125] The above is an illustrative scheme of an interactive playable ad automatic generation system based on structured analysis according to this embodiment. It should be noted that the technical solution of this interactive playable ad automatic generation system based on structured analysis belongs to the same concept as the technical solution of the interactive playable ad automatic generation method based on structured analysis described above. Details not described in detail in the technical solution of this interactive playable ad automatic generation system based on structured analysis can be found in the description of the technical solution of the interactive playable ad automatic generation method based on structured analysis described above.

[0126] In one embodiment of this application, a computing device is also provided, including a memory, a processor, and computer instructions stored in the memory and executable on the processor. When the processor executes the instructions, it implements the steps of the described method for automatically generating interactive playable advertisements based on structured analysis.

[0127] The above is an illustrative scheme of a computing device according to this embodiment. It should be noted that the technical solution of this computing device belongs to the same concept as the technical solution of the above-described method for automatically generating interactive playable advertisements based on structured analysis. For details not described in detail in the technical solution of the computing device, please refer to the description of the technical solution of the above-described method for automatically generating interactive playable advertisements based on structured analysis.

[0128] An embodiment of this application also provides a computer-readable storage medium storing computer instructions that, when executed by a processor, implement the steps of the aforementioned method for automatically generating interactive playable advertisements based on structured analysis.

[0129] The above is an illustrative scheme of a computer-readable storage medium according to this embodiment. It should be noted that the technical solution of this storage medium belongs to the same concept as the technical solution of the above-described method for automatically generating interactive playable advertisements based on structured analysis. For details not described in detail in the technical solution of the storage medium, please refer to the description of the technical solution of the above-described method for automatically generating interactive playable advertisements based on structured analysis.

[0130] The foregoing has described specific embodiments of this application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0131] The computer instructions include computer program code, which may be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium may be appropriately added to or subtracted according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media may not include electrical carrier signals and telecommunication signals.

[0132] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0133] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0134] The preferred embodiments disclosed above are merely illustrative of this application. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this application. These embodiments are selected and specifically described in this application to better explain the principles and practical applications of this application, thereby enabling those skilled in the art to better understand and utilize this application. This application is limited only by the claims and their full scope and equivalents.

Claims

1. A method for automatically generating interactive playable ads based on structured analysis, characterized in that, The method includes: The input video material is preprocessed to obtain a video frame sequence and a set of timing information text. Visual analysis is performed on the video frame sequence to obtain structured scene elements, and then the structured scene elements are fused with the temporal information text set to obtain semantic timeline labels; A game logic framework instance is generated based on the semantic timeline tags, and an interaction diagram is generated based on the structured scene elements and the game logic framework instance. Reconstruct interactive resource materials based on the structured scene elements described in the video; The interactive graph and the interactive resource materials are compiled and processed to generate an interactive trial advertisement product package.

2. The method according to claim 1, wherein, Preprocessing the input video footage yields a video frame sequence and a set of timing information text, including: The data in the input video material is parsed to extract the video frame sequence and independent audio stream; the video frame sequence is identified to obtain UI text with timestamps, and the audio stream is identified to obtain speech text with timestamps. The UI text and speech text with timestamps are used as a set of time-series information text.

3. The method according to claim 1, wherein, The structured scene elements obtained by visual analysis of the video frame sequence include: The video frame sequence is subjected to deep analysis to obtain structured scene elements in the video. The structured scene elements include: semantic segments, interactive objects, and key events / actions of interactive objects. Specifically, the video frame sequence is segmented into different semantic segments by detecting the shot boundaries. Interactive objects in a video frame sequence are identified through model recognition. By analyzing the movement trajectory and changes of interactive objects in a continuous video frame sequence, the key events / actions of the interactive objects can be obtained.

4. The method according to claim 3, wherein, The semantic timeline labels obtained by fusing the structured scene elements with the temporal information text set include: The semantic timeline labels are obtained by aligning the key events / actions of the interactive object with the text set containing time sequence information in time and then performing multimodal semantic fusion. The semantic timeline labels are used to mark the interactive atomic events that occur in the video.

5. The method according to claim 4, wherein, Generating game logic framework instances based on the semantic timeline labels includes: Feature vectors representing interaction patterns are extracted from the semantic timeline labels. The similarity between the extracted feature vectors and the preset feature vectors of each template in the preset gameplay template library is calculated. The template with the highest similarity is selected as the target template, and the set of configurable parameters in the target template is instantiated and filled.

6. The method according to claim 5, wherein, Generating an interaction diagram based on the structured scene elements and the game logic framework instance includes: The nodes of the interaction graph are constructed based on the target template and the semantic fragments; The edges of the interaction graph are constructed based on the target template and the key events / actions of the interactive object; Define specific actions on the nodes and edges to specify the specific operations that should be performed at a specific time; Serialize the interaction graph containing nodes, edges, actions, and their complete dependencies into a structured data format.

7. The method according to claim 3, wherein, The interactive resource materials reconstructed from the structured scene elements based on the video include: The identified interactive objects are extracted at the pixel level to obtain image resource materials; the motion trajectory of the image resource materials is estimated and fitted to generate motion description data of the image resource materials.

8. The method according to claim 7, wherein, The interactive graph and the interactive resource materials are compiled and processed to generate an interactive playable advertisement product package, including: The interaction graph is traversed and parsed, and the node, edge, and action definitions are translated into executable script instructions; a mapping index is established between the script instructions and the interactive resource materials. The interactive playable advertising product package is obtained by embedding the script instructions, the interactive resource materials, and the mapping index between the two into the software package along with a pre-built runtime engine.

9. An automatic interactive playable ad generation system based on structured analysis, characterized in that, include: The preprocessing unit is used to preprocess the input video material to obtain a video frame sequence and a set of timing information text. A visual analysis unit is used to perform visual analysis on the video frame sequence to obtain structured scene elements; The fusion unit is used to fuse the structured scene elements with the temporal information text set to obtain semantic timeline labels; The generation unit is used to generate a game logic framework instance based on the semantic timeline labels, and to generate an interaction diagram based on the structured scene elements and the game logic framework instance. A reconstruction unit is used to reconstruct interactive resource materials based on the structured scene elements in the video. The compilation unit is used to compile and process the interactive graph and the interactive resource materials to generate an interactive trial advertisement product package.

10. A computing device, comprising a memory, a processor, and computer instructions stored in the memory and executable on the processor, characterized in that, When the processor executes the instructions, it implements the steps of the method according to any one of claims 1-8.

11. A computer-readable storage medium storing computer instructions, characterized in that, When executed by the processor, this instruction implements the steps of the method according to any one of claims 1-8.

Images