Video processing method, automatic question answering method, computing device, computer readable storage medium, and computer program product

Abstract

Embodiments of the present disclosure provide a video processing method, an automatic question answering method, a computing device, a computer readable storage medium, and a computer program product. The video processing method comprises: obtaining a first initial video frame in an initial video and at least one second initial video frame corresponding to the first initial video frame; determining a reference video corresponding to the initial video, and determining a first reference video frame corresponding to the first initial video frame and a second reference video frame corresponding to each second initial video frame; on the basis of a preset motion feature generation policy, generating motion feature information between the first initial video frame and each second initial video frame; and processing the first reference video frame on the basis of each piece of motion feature information, and obtaining a target video frame corresponding to the first reference video frame and generating a target video on the basis of the at least one target video frame. The first video frame is processed on the basis of the motion feature information of various first video frames in the reference video, thereby rendering the reference video into a smooth target video.

Description

[0001]TECHNICAL FIELD Embodiments of the present disclosure relate to the field of computer technology, and more particularly to a video processing method. Background: In recent years, the field of image processing has experienced rapid development. In particular, the emergence of diffusion models trained on massive datasets has marked a major revolution in image synthesis technology. These diffusion models not only surpass generative adversarial networks in quality, but have also achieved remarkable results in areas such as image style transfer, super-resolution, and image editing. However, when applying these advanced image processing technologies to video processing, maintaining video consistency is a challenge. In particular, in tasks such as video style transfer, since each frame of a video is processed independently, this often leads to incoherent content between frames, resulting in flickering in the generated video. Therefore, there is an urgent need for a video processing method that can effectively address video consistency issues. SUMMARY OF THE INVENTION In view of this, embodiments of the present disclosure provide a video processing method, an automatic question answering method, and a video processing method for a cloud server. One or more embodiments of the present disclosure relate to a video processing apparatus, a computing device, a computer-readable storage medium, and a computer program product to address technical deficiencies in the prior art. According to a first aspect of an embodiment of the present disclosure, a video processing method is provided, comprising: obtaining a first initial video frame in an initial video and at least one second initial video frame corresponding to the first initial video frame; determining a reference video corresponding to the initial video, and determining a first reference video frame corresponding to the first initial video frame and second reference video frames corresponding to each second initial video frame, wherein the reference video has the same number of video frames as the initial video; generating motion feature information between the first initial video frame and each second initial video frame based on a preset motion feature generation strategy; processing the first reference video frame based on the motion feature information to obtain a target video frame corresponding to the first reference video frame, and generating a target video based on the at least one target video frame.According to a second aspect of an embodiment of the present disclosure, a video processing method applied to a cloud server is provided, comprising: receiving a video processing instruction sent by a terminal GAL 1 device, wherein the video processing instruction carries an initial video and a reference video corresponding to the initial video; obtaining a first initial video frame in the initial video, and at least one second initial video frame corresponding to the first initial video frame; obtaining a reference video corresponding to the initial video, and determining a first reference video frame corresponding to the first initial video frame and a second reference video frame corresponding to each second initial video frame, wherein the number of video frames of the reference video is the same as the number of video frames of the initial video; generating motion feature information between the first initial video frame and each second initial video frame based on a preset motion feature generation strategy; processing the first reference video frame according to each motion feature information, obtaining a target video frame corresponding to the first reference video frame, and generating a target video based on at least one target video frame, and returning the target video to the terminal GAL 1 device. According to a third aspect of an embodiment of the present disclosure, an automatic question-answering method is provided, comprising: receiving a video modification instruction, wherein the video modification instruction carries an initial video and an initial question text; parsing the video modification instruction to determine the initial question text, and generating an initial video modification prompt text based on the initial question text; inputting the initial video and the initial video modification prompt text into a video modification model to obtain a reference video generated by the video modification model; obtaining motion feature information between a target first video frame and each second video frame in the reference video based on the initial video, wherein the target first video frame is any one video frame in the reference video, and the second video frame is at least one video frame corresponding to the target first video frame in the reference video; adjusting each target first video frame in the reference video to generate a target video based on the motion feature information corresponding to each target first video frame in the reference video, and generating video modification answer information based on the target video and the initial question text. According to a fourth aspect of an embodiment of the present disclosure, a computing device is provided, comprising: a memory and a processor; the memory is configured to store computer-executable instructions, and the processor is configured to execute the computer-executable instructions. When executed by the processor, the computer-executable instructions implement the steps of the aforementioned video processing method, an automatic question-answering method, or a video processing method applied to a cloud server. According to a fifth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided, the medium storing computer-executable instructions. When executed by the processor, the instructions implement the steps of the aforementioned video processing method, an automatic question-answering method, or a video processing method applied to a cloud server.According to a sixth aspect of the embodiments of the present disclosure, a computer program product is provided, comprising a computer program / instructions. When executed by a processor, the computer program / instructions implement the steps of the aforementioned video processing method, an automatic question-answering method, or a video processing method applied to a cloud server. One embodiment of the present disclosure implements the steps of obtaining a first initial video frame and at least one second initial video frame corresponding to the first initial video frame from an initial video; determining a reference video corresponding to the initial video, and determining a first reference video frame corresponding to the first initial video frame and second reference video frames corresponding to each second initial video frame, wherein the reference video has the same number of video frames as the initial video; generating motion feature information between the first initial video frame and each second initial video frame based on a preset motion feature generation strategy; processing the first reference video frame based on the motion feature information to obtain a target video frame corresponding to the first reference video frame; and generating a target video based on the at least one target video frame. By applying the solution of the embodiments of the present disclosure, the motion feature information between each frame in the reference video determined by the motion feature information between video frames in the initial video can effectively reflect the motion feature information between each frame in the reference video. Subsequently, the motion feature information between each frame in the reference video determined by the above calculation of the feature information in the initial video can determine the expected motion relationship between each frame in the reference video based on the smooth initial video. Furthermore, the motion feature information between each frame in the smooth video can be used to process the non-smooth reference video, thereby avoiding the video flickering problem caused by discontinuity between frames, and thus achieving the rendering of the flickering reference video into a smooth target video.BRIEF DESCRIPTION OF THE DRAWINGS FIG1 is a flowchart of a video processing method provided by an embodiment of the present disclosure; FIG2a is a schematic diagram of a process mixing table of video frames and left Gamma-1 video frames in a video processing method provided by an embodiment of the present disclosure; FIG2b is a schematic diagram of a process mixing table of video frames and right Gamma-1 video frames in a video processing method provided by an embodiment of the present disclosure; FIG3a is a schematic diagram of a result mixing table of video frames and left Gamma-1 video frames in a video processing method provided by an embodiment of the present disclosure; FIG3b is a schematic diagram of a result mixing table of video frames and right Gamma-1 video frames in a video processing method provided by an embodiment of the present disclosure; FIG4 is a flowchart of a video processing method applied to a cloud server provided by an embodiment of the present disclosure; FIG5 is an architecture diagram of a video processing system provided by an embodiment of the present disclosure; FIG6 is a flowchart of an automatic question-answering method provided by an embodiment of the present disclosure; FIG7 is a process flow chart of a video optimization method provided by an embodiment of the present disclosure; FIG8 is a schematic diagram of the structure of a video processing device provided by an embodiment of the present disclosure; FIG9 is a block diagram of the structure of a computing device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION The following description sets forth numerous specific details to facilitate a thorough understanding of the present disclosure. However, the present disclosure can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without departing from the scope of the present disclosure. Therefore, the present disclosure is not limited to the specific embodiments disclosed below. The terminology used in one or more embodiments of the present disclosure is for the purpose of describing specific embodiments only and is not intended to limit the present disclosure. As used in one or more embodiments of the present disclosure and the appended claims, the singular forms "a," "an," "the," and "the" are 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 the present disclosure refers to and encompasses any and all possible combinations of one or more of the associated listed items. It should be understood that while the terms "first," "second," and so on may be used to describe various information in one or more embodiments of the present disclosure, such information should not be limited to these terms. These terms are used solely to distinguish information of the same type from one another. For example, "first" could be referred to as "second," and similarly, "second" could be referred to as "first," without departing from the scope of one or more embodiments of the present disclosure. Depending on the context, the term "if" as used herein could be interpreted as "when," "when," or "in response to determining."Furthermore, it should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, storage, and display, etc.) involved in one or more embodiments of the present disclosure are all information and data authorized by the user or fully authorized by all parties. The collection, use, and processing of the relevant data must comply with the relevant laws, regulations, and standards of the relevant countries and regions, and corresponding operation portals are provided for users to choose to authorize or reject. In one or more embodiments of the present disclosure, a large model refers to a deep learning model with large-scale model parameters, typically containing hundreds of millions, tens of billions, hundreds of billions, trillions, or even more than ten trillion model parameters. Large models can also be called cornerstone models / foundation models. They are pre-trained using large-scale unlabeled corpora to produce pre-trained models with parameters exceeding 100 million. Such models are adaptable to a wide range of downstream tasks and have good generalization capabilities. Examples include large language models (LLMs) and multi-modal pre-training models. In practical applications, large models only require a small number of samples to fine-tune the pre-trained model and can be applied to various tasks. Large models can be widely used in fields such as natural language processing (NLP) and computer vision. Specifically, they can be applied to computer vision tasks such as visual question answering (VQA), image captioning (IC), and image generation, as well as natural language processing tasks such as text-based sentiment classification, text summarization, and machine translation. Key application scenarios for large models include digital assistants, intelligent robots, search, online education, office software, e-commerce, and intelligent design. First, the terms used in one or more embodiments of this disclosure are explained. Match backpropagation leverages the spatial continuity between image blocks. Once the algorithm finds a matching region for an image block, it assumes that adjacent image blocks may also have a good match at a similar location in the target image. Therefore, this matching information is "backpropagated" to the adjacent blocks as an initial guess for their matching regions. This greatly speeds up the matching process by reducing the scope and number of searches. Random search: Once a preliminary matching area is determined for an image block, the algorithm will perform a random search around the match to find a better match.Random search is performed within a shrinking window, which allows for gradual refinement of matching results while maintaining high efficiency. This strategy assumes that matching blocks are typically located near the current matching block. Diffusion models trained on massive datasets have ushered in a revolutionary era in image synthesis. Diffusion models have proven to comprehensively outperform generative adversarial networks, even reaching a level of creative prowess comparable to that of human artists. However, extending these image processing techniques to video processing presents the challenge of maintaining smooth video. In particular, in video style transfer, since each frame in a video is processed independently, directly applying image processing methods often results in incoherent content, resulting in noticeable flicker in the generated video. To address this flickering issue, numerous methods have been proposed to enhance the consistency of the generated video and avoid the video artifact. For example, full video rendering methods process each frame through a deep learning model. To enhance frame consistency, mechanisms specifically designed for video processing are needed. Key frame sequence rendering methods use deep learning models to process key video frame sequences while simultaneously generating the remaining frames using interpolation methods. Single-frame rendering methods use deep learning models to process only a single frame, then render the complete video based on motion information extracted from the original video. However, these methods still have drawbacks. Existing methods struggle to ensure video coherence in full-video rendering, and in some cases, noticeable flickering can still occur. In key frame sequence rendering methods, the content of adjacent key frames can remain inconsistent, resulting in abrupt transitions in the rendered video. In single-frame rendering methods, due to the limited information available in a single frame, frame tearing often occurs in high-speed motion videos. This disclosure provides a video processing method, an automatic question-answering method, and a video processing method for a cloud server. This disclosure also relates to a video processing apparatus, a computing device, a computer-readable storage medium, and a computer program product, each of which is described in detail in the following embodiments. Typically, this video processing method is applied in scenarios where a user uploads an original video to a style transfer service. The style transfer service provider processes the uploaded original video using a deep learning model to generate a reference video that is similar to the original video but has a different style. Because the reference video directly output by the deep learning model may lack frame coherence, the reference video directly output by the deep learning model may exhibit flickering. Therefore, the video processing method provided in this disclosure is applied to the video directly output by the deep learning model to ensure frame coherence and avoid flickering in the video after style transfer.Referring to Figure 1, a flowchart of a video processing method provided according to an embodiment of the present disclosure is shown, specifically including the following steps. Step 102: Obtain a first initial video frame in an initial video, and at least one second initial video frame corresponding to the first initial video frame. In practical applications, the initial video is the video before style conversion, the first initial video frame is any frame in the initial video, and the second initial video frame is a video frame near the first initial video frame. Specifically, the initial video can be understood as a video uploaded by a user. The user converts the uploaded initial video into a target video of a different style using a style conversion service, for example, converting an original real-life landscape video into an animated-style landscape video. The target video is the video after the initial video has undergone style conversion. It can be understood as a video with the target style specified by the user and the same content as the initial video. During the conversion of the user-uploaded initial video into the target video, it is first processed by a deep learning model (preferably a large model such as a diffusion model) to obtain a reference video directly output by the deep learning model. The reference video is a video generated by the deep learning model based on the initial video. The reference video can be understood as the video after the initial video has undergone style conversion. Because the video frames in the reference video may not be coherent, the reference video is processed using the video processing method provided by the present disclosure to obtain a target video with coherent frames, and the target video is presented to the user to avoid directly presenting an incoherent video to the user. It should be noted that the first initial video frame is any frame in the initial video, and determining the second initial video frame corresponding to the first initial video frame can be understood as determining a preset number of frames near the first initial video frame. Specifically, determining the second initial video frame corresponding to the first initial video frame is preferably performed using a sliding window with a preset step size. The video frame at the center of the sliding window is the first initial video frame, and the remaining frames in the sliding window are the second initial video frames. It should be noted that if, after the sliding window slides to the target video frame, the number of video frames surrounding the target video frame does not meet the preset step size, only the video frames meeting the step size are selected.In one embodiment provided herein, a user uploads an initial video, Video A, which has three video frames: Initial Video Frame 0, Initial Video Frame 1, and Initial Video Frame 2. When the preset step size is 1, a sliding window acquires a video frame before and after the target video frame. The sliding window begins sliding, and the initial position of the sliding window is Initial Video Frame 0. The current first initial video frame is confirmed to be Initial Video Frame 0, and the second initial video frame is a video frame adjacent to Initial Video Frame 0, namely, Initial Video Frame 1. Subsequently, when the sliding window slides to Initial Video Frame 1, the current first initial video frame is confirmed to be Initial Video Frame 1, and the second initial video frame is a video frame adjacent to Initial Video Frame 1, namely, Initial Video Frame 0 and Initial Video Frame 2. Subsequently, when the sliding window slides to Initial Video Frame 2, the current first initial video frame is confirmed to be Initial Video Frame 2, and the second initial video frame is a video frame adjacent to Initial Video Frame 2, namely, Initial Video Frame 1. The sliding window is then confirmed to have reached the last video frame of the video, and the sliding of the sliding window ends. By acquiring a first initial video frame from an initial video and second initial video frames surrounding the first initial video frame, it is possible to subsequently acquire the association relationship between the various video frames in a coherent initial video. This association relationship between the various frames in the acquired initial video frames can then be used to process an incoherent reference video to generate a coherent target video. Step 104: Determine a reference video corresponding to the initial video, and determine a first reference video frame corresponding to the first initial video frame and second reference video frames corresponding to each second initial video frame. The reference video has the same number of video frames as the initial video. In practical applications, the first reference video frame is the video frame at the same position in the reference video as the first initial video frame, and the second reference video frame is the video frame at the same position in the reference video as the second initial video frame. Specifically, in a solution for optimizing a reference video output by a deep learning model when the reference video has the same number of frames as the initial video, since the reference video has the same number of frames as the initial video, reconfirming the first reference video frame relative to the first initial video frame may be determining a video frame at the same position as the first initial video frame as the first reference video frame. Similarly, confirming the second reference video frame relative to the second initial video frame may be determining a video frame at the same position as the second initial video frame as the second reference video frame.In one embodiment provided herein, a user uploads an initial video A consisting of three video frames: initial video frame 0, initial video frame 1, and initial video frame 2. Subsequently, a deep learning model generates an anime-style reference video B based on initial video A. Reference video B also consists of three video frames: reference video frame 0, reference video frame 1, and reference video frame 2. Continuing with the above example, it is determined that there are a first initial video frame, initial video frame 1, and second initial video frames, initial video frame 0 and initial video frame 2. Since initial video frame 1 is the second frame in the initial video, the first reference video frame corresponding to the first initial video frame is the second frame in the reference video, namely, reference video frame 1. Similarly, since initial video frame 0 and initial video frame 2 are the first and third frames in the initial video, respectively, the second reference video frames corresponding to the second initial video frame are the first and third frames in the reference video, namely, reference video frame 0 and reference video frame 2. By identifying the first and second reference video frames corresponding to the first and second initial video frames in the reference video, it is possible to subsequently process the incoherent reference video by identifying the associations between the frames in the acquired initial video frames to generate a coherent target video. Step 106: Based on a preset motion feature generation strategy, motion feature information is generated between the first initial video frame and each second initial video frame. In practical applications, the motion feature generation strategy is an information generation strategy that determines the connection information between the first initial video frame and the surrounding second initial video frames, and the motion feature information is the connection information between the first initial video frame and the surrounding second initial video frames. Specifically, the motion feature information between the first initial video frame and each second initial video frame can be understood as the corresponding position of each pixel in the second initial video frame in the first initial video frame when each second initial video frame changes to the first initial video frame. This information also reflects the change characteristics of each second initial video frame from the first initial video frame to the first initial video frame. Furthermore, it can be further understood as the motion characteristics representing the change of an entity in each second initial video frame to the corresponding entity in the first initial video frame. By calculating the motion feature information between frames in the reference video and determining it through the aforementioned calculation of the feature information in the initial video, the desired motion relationship between frames in the reference video can be determined from the smooth initial video. Furthermore, the non-smooth reference video can be processed using the motion feature information between frames in the smooth initial video, thereby avoiding the video flickering problem caused by discontinuities between frames and rendering the flickering reference video into a smooth target video.Considering that obtaining motion feature information between video frames is the change relationship between each pixel between two video frames, the video frame includes at least one pixel point; further, based on a preset motion feature generation strategy, generating motion feature information between the first initial video frame and each second initial video frame includes: determining a target second initial video frame, and determining at least one intermediate pixel point from the target second initial video frame, wherein the target second initial video frame is any one of the initial second video frames; obtaining initial pixel motion feature information, and adjusting the initial pixel motion feature information according to each intermediate pixel point and the first initial video frame to obtain target pixel motion feature information corresponding to each intermediate pixel point; and determining motion feature information between the first initial video frame and the target second initial video frame according to each target pixel motion feature information. In practical applications, the intermediate pixels are pixels in the target second initial video frame, the initial pixel motion feature information is randomly initialized pixel motion feature information, and the target pixel motion feature information is adjusted pixel motion feature information. Specifically, the target pixel motion feature information is the position of each intermediate pixel in the target second initial video frame corresponding to a pixel in the first initial video frame. By determining the corresponding position of each pixel in the target second initial video frame in the first initial video frame, the motion feature information between the target second initial video frame and the first initial video frame can be determined. Furthermore, the non-smooth reference video can be processed based on the motion feature information between each frame in the smooth initial video, thereby avoiding video flicker caused by discontinuities between frames and rendering the flickering reference video into a smooth target video. Considering the correspondence between pixels and the change relationship between pixels, it is necessary to compare two pixels and calculate the difference to determine the other pixel that best matches a certain pixel. Therefore, the motion feature generation strategy includes a motion feature adjustment function. Furthermore, the initial pixel motion feature information is adjusted according to each intermediate pixel point and the first initial video frame, including: processing the intermediate pixel point according to the initial pixel motion feature information to obtain a reference pixel point; obtaining a pixel difference value corresponding to the reference pixel point according to the motion feature adjustment function, the intermediate pixel point, and the reference pixel point; adjusting the initial pixel motion feature information according to the pixel difference value to obtain reference pixel motion feature information; using the reference pixel motion feature information as the initial pixel motion feature information, and continuing to perform the step of processing the intermediate pixel point according to the initial pixel motion feature information to obtain the reference pixel point until the adjustment stop condition is met.In practical applications, the motion feature adjustment function is a function that adjusts the initial pixel motion feature information. The reference pixel is a pixel obtained by processing an intermediate pixel based on the initial pixel motion feature information. The pixel difference is the difference between the pixel obtained after processing according to the initial pixel motion feature and the actual corresponding pixel in the first initial video frame. The reference pixel motion feature information is the pixel motion feature information obtained after adjusting the initial pixel motion feature information. Specifically, the reference pixel can be understood as the pixel coordinate obtained by processing the coordinates of the intermediate pixel based on the initial pixel motion feature information. The pixel corresponding to the processed pixel coordinate in the first initial video frame is determined based on the processed pixel coordinate, which is the reference pixel. It should be noted that the pixel here includes coordinate data and color data. When subsequently calculating the pixel difference, it is preferably calculated as the pixel color difference. The specific method for adjusting the initial pixel motion feature information can be any method for obtaining pixels, such as matching feedback, random search, etc., and this disclosure does not impose any limitation on this. The adjustment stopping condition may be, for example, the aforementioned pixel difference reaching a target value, or the algorithm iteration number reaching a target number, and the present disclosure imposes no limitations thereon. Considering that the reference pixel motion feature information obtained after adjustment is stopped upon reaching the adjustment stopping condition is not necessarily the reference pixel motion feature information with the smallest pixel difference, it is preferred that, after adjustment is stopped, the pixel motion feature information corresponding to the reference pixel with the smallest pixel difference is set as the target pixel motion feature information. Taking into account that when processing a video, the efficiency of video processing can be improved on the basis of reducing computing power consumption, or the result of video processing can be further improved without considering computing power consumption, the motion feature adjustment function includes a motion feature area adjustment function or a motion feature overall adjustment function; further, according to the motion feature adjustment function, the intermediate pixel point and the reference pixel point, the pixel difference value corresponding to the reference pixel point is obtained, including: when the motion feature adjustment function is a motion feature area adjustment function, according to the reference pixel point and the intermediate pixel point, the pixel difference value is obtained; when the motion feature adjustment function is a motion feature overall adjustment function, the error pixel point corresponding to the intermediate pixel point in each reference second initial video frame is obtained, and the pixel difference value is obtained according to each error pixel point, the intermediate pixel point and the reference pixel point, wherein the reference second initial video frame is a second initial video frame other than the target second initial video frame.In practical applications, the initial pixel point is a pixel point obtained by using the coordinates of the reference pixel point in the first initial video frame. The motion feature region adjustment function is a calculation function that determines the pixel difference value based on the difference between the reference pixel point after the change and the intermediate pixel point before the change. The error pixel point is a pixel point corresponding to the intermediate pixel point in the other second initial video frame. The motion feature overall adjustment function is a calculation function that determines the pixel difference value based on the difference between the reference pixel point after the change and the intermediate pixel point before the change and the error pixel point in the other second initial video frame. Specifically, the error pixel point can be understood as the average value of the color of the pixel points with the same coordinates as the above-mentioned coordinates in the other second initial video frames excluding the target second initial video frame based on the coordinates of the intermediate pixel point, that is, the error pixel point. In one embodiment provided by the present application, the motion feature adjustment function for calculating the pixel difference value is the motion feature region adjustment function. The specific calculation method is shown in Formula (1): Wherein, x,y is the coordinate of the middle pixel point, S is the first initial video frame, T is the second initial video frame, F is the initial pixel motion feature information, &S, T, F)x,y is the pixel difference value of the pixel point with coordinates x, y, that is, the pixel difference value of the middle pixel point after processing according to the initial pixel motion feature information, S[F(x,y)] is the pixel point in the first initial video frame after the coordinates of the middle pixel point are processed by the initial pixel motion feature information, that is, the reference pixel point, T[x, y] is the pixel point in the target second initial video frame with coordinates of the middle pixel point, that is, the middle pixel point, wherein the calculation method is to calculate the color difference between the two pixels and square it. In another embodiment provided by the present application, the motion feature adjustment function for calculating the pixel difference value is the overall motion feature adjustment function, and the specific calculation method is shown in formula (2): Initial pixel motion feature information, P is the average value of the remaining second initial video frames, a is a preset weight parameter. The larger the parameter, the greater the weight of the target initial video frame, which means that during the adjustment process, more emphasis is placed on the similarity between the reference pixel and the intermediate pixel. £(S, T, F)x,y is the difference value of the pixel with coordinates , that is, the pixel difference value of the intermediate pixel after processing based on the initial pixel motion feature information. S[F(x,y)] is the pixel in the first initial video frame whose coordinates are processed by the initial pixel motion feature information, which is the reference pixel. T[x, y] is the pixel in the target second initial video frame whose coordinates are the coordinates of the intermediate pixel, which is the intermediate pixel. T[x, y] is the average color of the pixels with the same coordinates as the intermediate pixel in the remaining second initial video frames excluding the target second initial video frame. The calculation method is to calculate the color differences between the intermediate pixel and the reference pixel and the error pixel, square them, and perform a weighted sum of the calculation results. Considering that if the color of a pixel after the change is similar to that before the change, the probability that the changed pixel is the pixel after the change in the second initial video frame is greater, the target pixel motion characteristics of each pixel can be automatically obtained by adjusting the pixel change pattern (initial pixel motion characteristics) by comparing the intermediate pixels before the change and the transformed reference pixels. Step 108: Process the first reference video frame based on the motion characteristic information to obtain a target video frame corresponding to the first reference video frame, and generate a target video based on at least one target video frame. In practical applications, the target video frame is a video frame in the target video where the pixel differences between adjacent target video frames are smooth. Specifically, the target video frame can be understood as a first reference video frame in the reference video fused with the features of a nearby second reference video frame to reduce the difference between the first reference video frame and the nearby second reference video frame. This avoids video flicker caused by discontinuity between frames in the video, thereby rendering the flickering reference video into a smooth target video.Furthermore, processing the first reference video frame based on each motion feature information to obtain a target video frame corresponding to the first reference video frame includes: determining a target second reference video frame, wherein the target second reference video frame is any one of the second reference video frames; obtaining mapped motion feature information, wherein the mapped motion feature information is motion feature information between the first reference video frame and the target second reference video frame; processing the target second reference video frame based on the mapped motion feature information to generate a to-be-mixed video frame corresponding to the target second reference video frame; and obtaining a target video frame corresponding to the first reference video frame based on a preset video frame mixing rule, the first reference video frame, and each to-be-mixed video frame. In practical applications, the mapped motion feature information is motion feature information between a target second initial video frame corresponding to the target second reference video frame and the first initial video frame, the to-be-mixed video frame is a video frame resulting from a change in the second reference video frame based on the mapped motion feature information, and the video frame mixing rule is a rule for mixing the to-be-mixed video frames to generate the target video frame. Specifically, since the second reference video frame in the reference video corresponds to the second initial video frame at the same video frame position in the initial video, and the first reference video frame in the reference video corresponds to the first initial video frame at the same video frame position in the initial video, the motion feature information between the first reference video frame and the target second reference video frame is the motion information between the first initial video frame and the target second reference video frame. A video frame blending rule can be understood as a blending rule for blending video frames in a certain order. The specific blending method between video frames can be any method for fusing two video frames. For example, since the order and number of blending operations affect the blending result, a preset video frame blending rule can effectively blend the transformed video frames to be blended with the first reference video frame, so that the blended target frame is generally similar to the first reference video frame and the differences between the acquired target video frames are smooth. Furthermore, processing the target second reference video frame according to the mapped motion feature information to generate a to-be-mixed video frame corresponding to the target second reference video frame includes: determining at least one to-be-processed pixel point and a to-be-processed pixel region corresponding to each to-be-processed pixel point from the target second reference video frame, wherein the pixel region includes at least one pixel point; processing each to-be-processed pixel region according to the mapped motion feature information, determining a target pixel point corresponding to each to-be-processed pixel point, and determining a to-be-mixed video frame corresponding to the target second reference video frame according to each target pixel point.In practical applications, the pixel to be processed is a pixel in the target second reference video frame, the pixel region to be processed is the region consisting of all pixels within a preset distance around the pixel to be processed, and the target pixel is the pixel obtained by processing each pixel in the pixel region to be processed based on the mapped motion feature information. Specifically, determining the pixel region to be processed corresponding to the pixel to be processed can be understood as obtaining all pixels within a preset range of the pixel to be processed based on a preset distance. For example, if the coordinates of the pixel to be processed are (8, 9) and the preset range is 1, the obtained pixel region to be processed is the pixel region with a lower left corner at (7, 8) and an upper right corner at (9, 10). It should be noted that, considering that the region where the pixel to be processed resides may be at the edge of the video frame or any region to be processed within the preset distance cannot be obtained, black pixels are used to expand the video frame outward by the preset distance. Since the value of black pixels is 0 in the computer storage, using black pixels for expansion does not affect the pixels within the calculated region. After all pixels in the pixel region to be processed are processed using the corresponding mapped motion feature information, the target pixel can be obtained by processing all pixels in the pixel region to obtain the color corresponding to each pixel. The colors corresponding to the pixels are then averaged as the color of the target pixel. The coordinates of the pixel to be processed, obtained after processing the mapped motion feature information, are used as the coordinates of the target pixel to obtain the target pixel. By generating the replacement frame by modifying all pixels in a region corresponding to each pixel in each second reference video frame, this avoids the problem of large pixel disparity in the converted video frame caused by a one-to-one conversion based on individual pixels in the video frame. Furthermore, compared to directly modifying the entire video frame, segmenting the video frame into multiple regions and modifying each of the regions can reduce the problem of excessive video memory usage during the modification of the video frame, thereby improving video processing efficiency.Considering that in practical applications, video frame blending strategies can be stored in a special table to enable faster blending of video frames with less computing power during video processing, the video frame blending strategy includes a blending table query strategy. Furthermore, based on a preset video frame blending rule, the first reference video frame, and each to-be-blended video frame, obtaining a target video frame corresponding to the first reference video frame includes: constructing a blending table based on each first reference video frame and each to-be-blended video frame corresponding to each first reference video frame; querying the blending table for at least one intermediate blending result corresponding to the first reference video frame based on the blending table query strategy; and blending the intermediate blending results to obtain the target video frame corresponding to the first reference video frame. In practical applications, the blending table stores intermediate blending results of each first reference video frame under various preset sliding window sizes, the intermediate blending results are blending results obtained after preliminary blending of each to-be-blended frame, and the blending table query strategy queries the blending table for the intermediate blending results of each first reference video frame. Specifically, a mixing table can be understood as a tree-like structure storing intermediate mixing results obtained by preliminarily mixing various to-be-mixed results at various sliding window sizes. In practical applications, a larger sliding window does not necessarily result in better mixing results. Therefore, when processing video, it is necessary to obtain video processing results for various sliding window sizes for comparison and identification of the most effective video processing results. Therefore, rather than performing mixing of the to-be-mixed video frames each time the sliding window is adjusted, the intermediate mixing results of the to-be-mixed video frames at each sliding window size are stored in a storage table. This allows the corresponding intermediate mixing results to be used for mixing when generating the target video, reducing the number of times the to-be-mixed video frames are mixed, thereby improving target video generation efficiency. Preferably, the mixing table includes a process mixing table and a result mixing table. If the mixing table is a process mixing table, the corresponding query strategy for the mixing table is the process mixing table query strategy; if the mixing table is a result mixing table, the corresponding query strategy for the mixing table is the result mixing table query strategy. The process mixing table can be understood as storing the intermediate mixing results of each first reference video frame under each preset sliding window size in a tree structure. The process mixing table query strategy can be understood as a strategy for determining the intermediate mixing results that need to be mixed for each first reference video frame based on the sliding window size selected by the user, as shown in Figures 2a and 2b, Figure 2a is a schematic diagram of a process mixing table of a video frame and a left-example video frame in a video processing method provided by an embodiment of the present disclosure, and Figure 2b is a schematic diagram of a process mixing table of a video frame and a right-example video frame in a video processing method provided by an embodiment of the present disclosure.The blending level in FIG2a and FIG2b is a preset sliding window size. When the blending level is 0, the corresponding sliding window size is 2 to the power of 0, that is, 1. When the blending level is 1, the sliding window size is 1 to the power of 1, that is, 2. Similarly, a blending level of 2 corresponds to a sliding window size of 4, and a blending level of 3 corresponds to a sliding window size of 8. The target video frame is the sequence number of the target video frame obtained after processing the first reference video frame, that is, the sequence number of the first reference video frame. In the embodiment shown in the figure, the video corresponds to 8 video frames. Taking the mixing process of mixing level 2 and target video frame 3 as an example, first obtain the intermediate mixing result of target video frame 3 and the left example video frame. It can be seen from the table in Figure 2a that it is necessary to obtain the leaf node corresponding to target frame 3 and the ancestor node of the leaf node in mixing level 1 and mixing level 2, that is, the intermediate mixing results S3, S2->S3 and (S0->S3)+(S1->S3); then obtain the intermediate mixing results S3, S3<-S4 and (S3<-S5)+(S3<-S6) of target video frame 3 and the right example video frame in the same way; then remove the repeated S3 and mix the above five intermediate mixing results to obtain target video frame 3, that is, only four mixing sets are needed to obtain target video frame 3, and to mix the video frame to be mixed obtained by the sliding window, it is necessary to mix the first reference video frame S3 with S2->S3, S0->S3, S1->S3, S3<-S4, S3<-S5, The six video frames to be mixed (S3<-S6) total seven frames, requiring six mixing cycles to obtain target video frame 3. The result mixing table can be understood as a tree-like structure storing further intermediate mixing results for each first reference video frame under various preset sliding window sizes. The result mixing table query strategy can be understood as a strategy for determining the intermediate mixing results required for each first reference video frame based on the sliding window size selected by the user. As shown in Figures 3a and 3b, Figure 3a is a schematic diagram of a result mixing table for video frames and left and right video frames in a video processing method provided by one embodiment of the present disclosure; Figure 3b is a schematic diagram of a result mixing table for video frames and right and left video frames in a video processing method provided by one embodiment of the present disclosure. The mixing levels and target video frames in Figures 3a and 3b are identical to the technical features described above in Figures 2a and 2b, and will not be further elaborated here. In the embodiment shown in the figures, the video corresponds to eight video frames.Taking the blending process of target video frame 3 at blending level 2 as an example, the intermediate blending result of target video frame 3 and the left-hand example video frame is first obtained. Querying the table in Figure 2a shows that the ancestor node of the leaf node corresponding to target frame 3 at blending level 2 must be obtained, namely, the intermediate blending result (SO->S3)+(S2->S3)+(S1->S3)+S3. Subsequently, the intermediate blending result of target video frame 3 and the right-hand example video frame is obtained in the same manner: S3+(S3<-S5)+(S3<-S4)+(S3<-S6). These two intermediate blending results are then blended to obtain the video frame blending result corresponding to target video frame 3. Generating the target video frame using the intermediate blending result pairs stored in the result blending table and the process blending table effectively reduces the number of blending operations required to generate the target video when the user adjusts the sliding window size, thereby improving video processing efficiency. Considering that in addition to performing similarity mixing on each frame to generate a new video, it is also possible to extract key video frames of the video to be processed and insert transition video frames between the key video frames to achieve a smooth video, before obtaining the first initial video frame in the initial video and at least one second initial video frame corresponding to the first initial video frame, the method further includes: obtaining at least one initial key video frame in the initial video, and determining a reference key video frame corresponding to the initial key video frame from a reference video corresponding to the initial video; generating key video frame feature information between each initial key video frame based on a preset key video frame feature generation strategy; obtaining a transition video frame corresponding to each reference key video frame based on each reference key video frame and the key video frame feature information corresponding to each reference key video frame; and generating a target video based on each reference key video frame and the transition video frame corresponding to each reference key video frame. In practical applications, the initial key video frame is any key video frame in the initial video. The reference key video frame is a key video frame in the reference video corresponding to the initial key video frame. The key video frame feature generation strategy is an information generation strategy for generating motion information of two key video frames and an intermediate transition video frame. The key video frame feature information is information representing motion features of the two key video frames and the intermediate transition video frame. The transition video frame is a non-key video frame between the two key video frames. Specifically, the initial key video frame can be understood as a video frame in the initial video in which an entity exhibits motion exceeding a threshold. The method for determining the initial key video frame can be any method for determining a key frame, and this disclosure does not impose any limitation thereto.The reference video for obtaining the reference key video frames is a video generated by a deep learning model based on the initial video. This reference video may include only the reference video key frames obtained by processing the initial key video frames in the initial video, or it may include reference video frames corresponding to all initial video frames in the initial video. This application imposes no restrictions on this. A transition video frame can be understood as a non-key video frame between the reference key video frames in the reference video. The transition video frame is generated by processing the two initial key video frames based on the key video frame feature information corresponding to the two initial key video frames to generate a transition video frame between the two initial key video frames. It should be noted that the key video feature information corresponding to the two key video frames may not be a single piece, but may include multiple pieces of key video feature information. If multiple pieces of key video feature information are included, at most the same number of transition video frames as the key video feature information can be generated between the two key video frames. Considering that obtaining key video frame feature information between key video frames is a possible change relationship between pixels between two key video frames, a video frame includes at least one pixel point. Generating key video frame feature information between initial key video frames based on a preset key video frame feature generation strategy includes: determining a first initial key video frame and a second initial key video frame, and determining a reference transition video frame based on the first and second initial key video frames, wherein the first initial key video frame is any one of the initial key video frames, and the second initial key video frame is a key video frame adjacent to the first initial key video frame; determining an initial transition pixel point in the reference transition video frame; obtaining initial pixel transition feature information, determining a first reference pixel point and a second reference pixel point based on the initial pixel transition feature information and the initial transition pixel point, and adjusting the initial pixel transition feature information based on the first reference pixel point, the second reference pixel point, and the initial transition pixel point to generate target pixel transition feature information corresponding to each of the first initial pixel point and the second reference pixel point; and generating key video frame feature information between the initial key video frames based on each target pixel transition feature information.In practical applications, the first initial key video frame is any key frame in the initial video, the second initial key video frame is a key video frame adjacent to the first initial key video frame, the reference transition video frame is a non-key frame between the two initial key video frames, the initial transition pixel is any pixel in the reference transition video frame, the first reference pixel is the pixel corresponding to the initial transition pixel in the first initial key video frame after processing based on the initial pixel transition feature information, the second reference pixel is the pixel corresponding to the initial transition pixel in the second initial key video frame after processing based on the initial pixel transition feature information, the initial pixel transition feature information is randomly initialized pixel transition feature information, and the target pixel transition feature information is the adjusted initial pixel transition feature information. Specifically, there may be multiple reference transition video frames. In the case of multiple reference transition video frames, multiple target pixel transition feature information is generated between the first initial key video frame and the second initial key video frame, so that a maximum number of non-key frames equal to the number of reference transition video frames are subsequently inserted between the two key video frames. Pixel transition feature information is similar to the aforementioned motion feature information; both represent the changes between two video frames. However, the pixel transition feature information here represents the changes between two initial key video frames and a reference transition video frame in between. Specifically, the output pixel transition feature information includes multiple groups, each containing both the change feature information with the first initial key video frame (first pixel transition feature information) and the change feature information with the second initial key video frame (second pixel transition feature information). Accordingly, when generating the target video, the reference key video frames corresponding to the two adjacent initial key video frames can be processed separately, and the two obtained results can be blended to obtain the intermediate non-key video frame (i.e., the transition video frame).Similarly, considering that the correspondence between pixels and the pixel change relationship requires comparing two pixels and calculating the difference to determine the pixel that best matches a certain pixel, the key video frame feature generation strategy includes a transition video frame adjustment function; determining a first reference pixel and a second reference pixel based on the initial pixel transition feature information and the initial transition pixel point, and adjusting the initial pixel transition feature information according to the first reference pixel point, the second reference pixel point, and the initial transition pixel point to generate target pixel transition feature information corresponding to each of the first initial pixel point and the second reference pixel point, including: processing the initial transition pixel point according to the initial pixel transition feature information to obtain a first reference pixel point and a second reference pixel point; obtaining a transition pixel point difference value corresponding to the reference pixel point according to the transition video frame adjustment function, the initial transition pixel point, the first reference pixel point, and the second reference pixel point; adjusting the initial pixel transition feature information according to the transition pixel point difference value to obtain reference pixel transition feature information; using the reference pixel transition feature information as the initial pixel transition feature information, and continuing to process the initial transition pixel point according to the initial pixel transition feature information. The step of obtaining a first reference pixel and a second reference pixel is performed until an adjustment stop condition is met. In practical applications, the transition video frame adjustment function includes a first key video frame adjustment function and a second key video adjustment function, and the transition pixel difference value includes a first transition pixel difference value and a second transition pixel difference value. The first key video frame adjustment function is a function that calculates a first transition pixel difference value between a first reference pixel and an initial transition pixel point, where the first transition pixel difference value is a pixel difference value used to adjust the change characteristic information between the first initial key video frame. The second key video frame adjustment function is a function that calculates a second transition pixel difference value between a second reference pixel point and an initial transition pixel point, where the second transition pixel difference value is a pixel difference value used to adjust the change characteristic information between the second initial key video frame. It should be noted that the coordinate data and color data are included. When subsequently calculating pixel differences, preferably the pixel color differences are calculated. The specific method for adjusting the initial pixel transition feature information can be any method for obtaining pixels, such as matching return, random search, etc. In addition, the specific method for adjusting the initial pixel transition feature information can also include obtaining pixels in adjacent key video frames, which is not limited in this disclosure. In one embodiment provided in this application, the first key video frame adjustment function and the second key video frame are used. £(S, T, F r \ iyis the difference of the second transitional pixel point with coordinates %, y, and Si Tian (%, y)] is the pixel point obtained by processing the coordinates of the initial transitional pixel point with the initial pixel transition feature information in the first initial key video frame, which is the first reference pixel point, S r[ / y(x, y)] represents the coordinates of the initial transition pixel in the second initial key video frame after processing the initial pixel transition feature information, which is the second reference pixel. T[x, y] represents the coordinates of the pixel in the reference transition video frame whose coordinates are the coordinates of the intermediate pixel, which is the initial transition pixel. The difference between the first transition pixel is calculated by squaring the difference between the initial transition pixel and the first reference pixel, and then performing a weighted summation with the square of the difference between the first reference pixel and the second reference pixel. The difference between the second transition pixel is calculated by squaring the difference between the initial transition pixel and the second reference pixel, and then performing a weighted summation with the square of the difference between the first reference pixel and the second reference pixel. It should be noted that the adjustment stopping condition here can include, for example, that each of the two pixel differences mentioned above reaches the target value, or that the number of iterations of the algorithm reaches the target number, and this is not limited in any way by this disclosure. Considering that the reference pixel transition characteristic information obtained after adjustment is stopped upon reaching the adjustment stop condition is not necessarily the reference pixel transition characteristic information with the smallest pixel difference value, it is preferred that, after adjustment stops, the pixel transition characteristic information with the smallest sum of the two pixel differences is set as the target pixel transition characteristic information. By setting a function to calculate the reference frame and two key frames to obtain the final target pixel transition characteristic information, the accuracy of the obtained target pixel transition information can be improved, thereby enabling the insertion of smooth transition video frames between the key video frames of the reference video, thereby rendering a flickering reference video into a smoother target video. Considering that the rendered target video needs to be sent to the user to present the rendering results to the user, after generating the target video, the method further includes: sending the target video to the target user. In actual applications, the target user is the user who receives the target video. Specifically, the target user is typically the user who sends the initial video. The user sends the initial video and video modification information used to edit the initial video to the Yunjia 1 device that provides the video processing method. The Yunjia 1 device then generates a reference video similar to the first video and conforming to the video modification information based on the initial video and video modification information sent by the user. The Yunjia 1 device then performs the aforementioned steps on the video frames in the reference video to generate the target video. The target video can then be sent to the target user so that the target user can view the modified results of the initial video. The target user can also be a different user from the user who sent the initial video, and this disclosure does not impose any restrictions on this.Considering that users may make further adjustments to the target video after receiving it, after sending the target video to the target user, the method further includes: receiving video adjustment information sent by the target user regarding the target video; adjusting the target video based on the video adjustment information to obtain the adjusted target video; and returning the adjusted target video to the target user. In practical applications, the video adjustment information is information sent by the user to adjust the target video. The target video can be modified to generate a pre-processed video, which has not yet been processed in a standardized manner. The pre-processed video can then be processed using the above steps to generate the adjusted target video. The adjusted target video can be a video of a different style from the target video, or a video with an entity replaced in the target video. Alternatively, the target video can be generated by regenerating a video of the same style or type without modifying the target video, and this application does not impose any restrictions on this. When the video adjustment information sent by the user is to regenerate a video of the same style or type, the target video can be processed according to the same prompt used by the user to generate a reference video by processing the initial video, generating a reference video of the same style or type as the target video. The reference video can then be processed again to generate a smooth, adjusted target video with the same style or type as the previous target video. Alternatively, the prompt used to generate the reference video by processing the initial video can be obtained, and the initial video can be processed again according to the prompt to generate a reference video of the same style or type as the target video. The reference video can then be processed again to generate a smooth, adjusted target video with the same style or type as the previous target video, and so on. This disclosure does not impose any limitations on this. In one embodiment provided by the present disclosure, the target video is a video of a white rabbit eating grass, and the video adjustment information sent by the user is "Replace the white rabbit with a black rabbit." The target video is then adjusted according to the adjustment information to generate an adjusted target video of a black rabbit eating grass, and the adjusted target video is returned to the user. In another embodiment provided by the present disclosure, the target video is a real video of a white rabbit eating grass. The video adjustment information sent by the user is "Change this video to an anime style." The target video is then adjusted according to the adjustment information to generate an adjusted target video of a white rabbit eating grass in an anime style. The adjusted target video is then returned to the user.In another embodiment provided by the present disclosure, a target video is a real video of a white rabbit eating grass. The user sends video adjustment information such as "I don't think this video is good. Give me another similar video." The target video is then adjusted based on the adjustment information to generate another adjusted target video of a white rabbit eating grass. The adjusted target video is then returned to the user. By receiving the video adjustment information provided by the user based on the target video, adjusting the target video, and then returning the adjusted target video to the user, the target video can be further adjusted based on the user's needs. After returning the adjusted target video to the user, the user can adjust the target video again. Further adjustments can be made to the target video based on the adjustment information sent by the user until the user stops adjusting the target video. This ensures that the target video received by the user better meets the user's needs, thereby improving the user experience. Considering that adding certain prompt interactions when returning a video to a user can improve the user experience, after generating the target video, the method further includes: generating video adjustment prompt information based on the target video; sending the target video and the video adjustment prompt information to the target user; receiving video adjustment information sent by the target user based on the video adjustment prompt information; adjusting the target video according to the video adjustment information to obtain the adjusted target video, and returning the adjusted target video to the target user. In practical applications, the video adjustment prompt information is information that prompts the user to adjust the target video. Specifically, the video adjustment prompt information can be understood as information that prompts the user to perform common adjustments to the target video. For example, after returning the target video to the target user, at least one interactive button is generated in the interactive interface, such as Button 1: "Generate another video of the same type" or Button 2: "Generate a video of a different style." The user can click the corresponding button to generate the video adjustment information. In addition, the user can also edit the video adjustment information to adjust the target video, and so on. This disclosure does not impose any restrictions on this. By generating video adjustment prompts, users can adjust the target video based on these prompts. This provides users with video adjustment suggestions during interaction, thereby improving the user experience. The solution of the disclosed embodiments utilizes a motion feature generation strategy that does not require a deep learning model or a machine learning model. By calculating motion feature information between frames in the initial video, this reduces computing power consumption during video processing while ensuring the smoothness of the generated video.Furthermore, by calculating the motion feature information between frames in the reference video, the motion relationship between frames in the smooth initial video can be determined. This motion feature information can then be used to process the non-smooth reference video, avoiding the problem of video flickering caused by discontinuities between frames. This allows the flickering reference video to be rendered as a smooth target video. Furthermore, a special tree-like blending table stores the blending rules required for each frame. This allows for more efficient video processing, reducing time and computing power consumption and improving processing efficiency. Furthermore, a key video frame feature generation strategy is used to process key video frames in the smooth initial video to obtain features of transition frames between key video frames in the smooth video. This allows key video frame extraction from the non-smooth reference video. Key video frames of the reference video are obtained and processed, and smooth transition frames are inserted between the key video frames of the reference video, allowing the flickering reference video to be rendered as a smoother target video. Referring to Figure 4, a flowchart of a video processing method for a cloud server according to an embodiment of the present disclosure is shown. The method specifically includes the following steps: Step 402: Receive a video processing instruction sent by a Gigabit 1 device, wherein the video processing instruction carries an initial video and a reference video corresponding to the initial video. Step 404: Obtain a first initial video frame in the initial video and at least one second initial video frame corresponding to the first initial video frame. Step 406: Obtain a reference video corresponding to the initial video and determine a first reference video frame corresponding to the first initial video frame and second reference video frames corresponding to each second initial video frame, wherein the number of video frames in the reference video is the same as the number of video frames in the initial video. Step 408: Generate motion feature information between the first initial video frame and each second initial video frame based on a preset motion feature generation strategy. Step 410: Process the first reference video frame based on the motion feature information to obtain a target video frame corresponding to the first reference video frame, generate a target video based on the at least one target video frame, and return the target video to the Gigabit 1 device. The above is an illustrative embodiment of the video processing method for a cloud server. It should be noted that the technical solution of this video processing method for a cloud server shares the same concept as the technical solution of the aforementioned video processing method. For details not described in detail in the technical solution of the video processing method for a cloud server, please refer to the description of the technical solution of the aforementioned video processing method.The solution of the disclosed embodiments utilizes a motion feature generation strategy that does not require a deep learning model or a machine learning model. By calculating the motion feature information between frames in the initial video and determining it, the computational power consumption during video processing can be reduced while ensuring the smoothness of the generated video. Furthermore, by calculating the motion feature information between frames in the reference video, the motion relationship between frames in the reference video can be determined from the smooth initial video. Furthermore, the motion feature information between frames in the smooth video can be used to process the non-smooth reference video, avoiding the problem of video flicker caused by discontinuities between frames and rendering the flickering reference video into a smooth target video. Furthermore, a special tree-like blending table stores the blending rules required for each frame. This allows for more efficient video processing, reducing time and computing power consumption, and improving video processing efficiency. Referring to Figure 5, it shows an architecture diagram of a video processing system provided in accordance with an embodiment of the present disclosure. The video processing system may include a client 100 and a server 200. The client 100 is configured to send a video processing instruction carrying an initial video and a reference video corresponding to the initial video to the server 200. The server 200 is configured to obtain a first initial video frame from the initial video and at least one second initial video frame corresponding to the first initial video frame. The server 200 is configured to obtain a reference video corresponding to the initial video and determine a first reference video frame corresponding to the first initial video frame and second reference video frames corresponding to each second initial video frame, where the number of video frames in the reference video is the same as the number of video frames in the initial video. Motion feature information is generated between the first initial video frame and each second initial video frame based on a preset motion feature generation strategy. The server 200 processes the first reference video frame based on the motion feature information to obtain a target video frame corresponding to the first reference video frame, and generates a target video based on the at least one target video frame. The server 200 then sends the target video to the client 100. The client 100 is further configured to receive the target video sent by the server 200.Using the solution of the embodiments of the present disclosure, the motion feature information between frames in the reference video, determined based on the motion feature information between frames in the initial video, can effectively reflect the motion feature information between frames in the reference video. Subsequently, the motion feature information between frames in the reference video, determined by calculating the feature information in the initial video, can be used to determine the expected motion relationship between frames in the reference video based on the smooth initial video. Furthermore, the motion feature information between frames in the smooth video can be used to process the non-smooth reference video, avoiding video flicker caused by discontinuities between frames and rendering the flickering reference video into a smooth target video. The video processing system may include multiple clients 100 and a server 200. The client 100 may be referred to as a client G1 device, and the server 200 may be referred to as a cloud G1 device. Multiple clients 100 can establish a communication connection through the server 200. In a video processing scenario, the server 200 is used to provide video processing services between the multiple clients 100. The multiple clients 100 can act as senders or receivers, communicating through the server 200. Users can interact with the server 200 through the client 100 to receive data from other clients 100 or send data to other clients 100. In a video processing scenario, a user can publish a data stream to the server 200 through the client 100. The server 200 generates a target video based on the data stream and pushes the target video to other clients with which communication has been established. The connection between the client 100 and the server 200 is established through a network. The network provides the medium for the communication link between the client 100 and the server 200. The network can include various connection types, such as wired or wireless communication links or fiber optic cables. The data transmitted by the client 100 may need to be encoded, transcoded, compressed, or otherwise processed before being published to the server 200. The client 100 may be a browser, an APP (Application), a web application such as an H5 (HyperText Markup Languages, version 5) application, a light application (also known as a mini-program, a lightweight application), or a cloud application. The client 100 may be developed based on a software development kit (SDK) for a corresponding service provided by the server 200, such as a real-time communication (RTC) SDK.The client 100 can be deployed in an electronic device and rely on the device or certain apps in the device to operate. For example, the electronic device may have a display screen and support information browsing, such as a personal mobile terminal such as a mobile phone, tablet computer, or personal computer. Various other applications may also be configured in the electronic device, such as human-computer interaction applications, model training applications, text processing applications, web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc. The server 200 may include servers that provide various services, such as servers that provide communication services to multiple clients, servers that support backend training for models used by clients, and servers that process data sent by clients. It should be noted that the server 200 can be implemented as a distributed server cluster consisting of multiple servers, or as a single server. The server can also be a server in a distributed system or a server integrated with blockchain. The server may also be a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms, or an intelligent cloud computing server or intelligent cloud host with artificial intelligence technology. It is worth noting that the video processing method provided in the embodiments of the present disclosure is generally executed by the server. However, in other embodiments of the present disclosure, a client may also have similar functions to the server and thereby execute the video processing method provided in the embodiments of the present disclosure. In other embodiments, the video processing method provided in the embodiments of the present disclosure may also be jointly executed by the client and the server. Referring to Figure 6, Figure 6 shows a flowchart of an automatic question-answering method provided in one embodiment of the present disclosure, specifically comprising the following steps: Step 602: Receive a video modification instruction, wherein the video modification instruction carries an initial video and an initial question text. Step 604: Parse the video modification instruction to determine the initial question text, and generate an initial video modification prompt text based on the initial question text. Step 606: Input the initial video and the initial video modification prompt text into a video modification model to obtain a reference video generated by the video modification model.Step 608: Motion feature information between a target first video frame and each second video frame in the reference video is obtained based on the initial video, where the target first video frame is any video frame in the reference video, and the second video frame is at least one video frame corresponding to the target first video frame in the reference video. Step 610: Adjust each target first video frame in the reference video based on the motion feature information corresponding to each target first video frame in the reference video to generate a target video, and generate video modification answer information based on the target video and the initial question text. In actual applications, the video modification instruction is an instruction to modify the initial video, the initial question text is text entered by the user to modify the initial video, the initial video modification prompt text is text input into the model for modifying the initial video, and the video modification model is a model that modifies the initial video based on the initial video modification prompt text. Specifically, a user sends a video modification instruction containing an initial video and an initial question text for editing the initial video to a cloud-based device providing a video processing method. The cloud-based device then generates an initial video modification prompt text based on the initial question text sent by the user. This prompt text and the initial video are input into a video modification model, which then obtains a reference video output by the video modification model. Because the reference video directly output has video discontinuities, each frame of the reference video is processed as described in the aforementioned video processing method to obtain a smoother target video. Video modification answer information is then generated based on the target video and returned to the user. The user can further modify the target video until they stop modifying the video. The video modification model can be understood as a deep learning model for video modification (preferably a large model such as a diffusion model). The video modification model can modify the style or content of the initial video based on the text sent by the user, and this disclosure does not impose any limitations on this. The above is an exemplary embodiment of the automatic question answering method of this embodiment. It should be noted that the technical solution of the automatic question answering method and the technical solution of the above-mentioned video processing method belong to the same concept. For details not described in detail in the technical solution of the automatic question answering method, please refer to the description of the technical solution of the above-mentioned video processing method.Using the solution of an embodiment of the present disclosure, a video modification instruction sent by a user is received, the initial video to be modified is modified, and a reference video modified according to the target video is generated. Then, using a motion feature generation strategy that does not require a deep learning model or a machine learning model, motion feature information is calculated between frames in the initial video to determine motion feature information. This ensures the smoothness of the generated video while reducing computing power consumption during video processing. Furthermore, by calculating the motion feature information between frames in the reference video, the motion relationship between frames in the reference video can be determined from the smooth initial video. Furthermore, the motion feature information between frames in the smooth video can be used to process the non-smooth reference video, avoiding the problem of video flicker caused by discontinuities between frames, thereby rendering the flickering reference video into a smooth target video. The following description further illustrates the video processing method provided by the present disclosure using the example of its application in video optimization, with reference to FIG7 . FIG7 shows a process flow chart of a video optimization method provided by an embodiment of the present disclosure, which specifically includes the following steps. Step 702: Obtain an original video and a reference video after style adjustment. Step 704: Set a sliding window with a step size of 2, query a tree-structured blending table, and determine a blending method for each target video frame in the reference video and other video frames within the sliding window. Step 706: Obtain motion feature information for the other video frames and the target video frame within the sliding window based on the determined blending method, and process the other video frames within the sliding window based on the acquired motion feature information. It should be noted that the specific method for obtaining the motion feature information for the other video frames and the target video frame within the sliding window is to determine the corresponding video frames in the original video for the target video frame and other video frames in the reference video, and then obtain the motion feature information for the other video frames and the target video frame based on the corresponding video frames in the original video. Step 708: Blend the processed other video frames and the target video frame according to the blending method to obtain the processed video frames. Step 710: Process all target video frames in the reference video to obtain multiple processed video frames, and splice the processed video frames into the target video. By applying the solution of the embodiments of the present disclosure, a motion feature generation strategy that does not require a deep learning model or a machine learning model is employed. By calculating the motion feature information between frames in the initial video, the computational power consumption during video processing can be reduced while ensuring the smoothness of the generated video.Furthermore, by calculating the motion feature information between frames in the reference video, the motion relationship between frames in the smooth initial video can be determined. Furthermore, the non-smooth reference video can be processed based on the motion feature information between frames in the smooth video. This avoids video flickering caused by discontinuities between frames, thereby rendering the flickering reference video into a smooth target video. Furthermore, a special tree-like blending table stores the blending rules required for each frame. This allows for more efficient video processing, reducing time and computing power consumption, and improving video processing efficiency. Corresponding to the above-described method embodiments, the present disclosure also provides embodiments of a video processing device. FIG8 shows a schematic structural diagram of a video processing device provided by one embodiment of the present disclosure. As shown in Figure 8, the device includes: an initial video frame acquisition module 802, configured to acquire a first initial video frame in an initial video, and at least one second initial video frame corresponding to the first initial video frame; a reference video frame confirmation module 804, configured to determine a reference video corresponding to the initial video, and determine a first reference video frame corresponding to the first initial video frame and a second reference video frame corresponding to each second initial video frame, wherein the number of video frames of the reference video is the same as the number of video frames of the initial video; a motion feature information generation module 806, configured to generate motion feature information between the first initial video frame and each second initial video frame based on a preset motion feature generation strategy; a target video generation module 808, configured to process the first reference video frame according to each motion feature information, obtain a target video frame corresponding to the first reference video frame, and generate a target video based on at least one target video frame. Optionally, the video frame includes at least one pixel point; the motion feature information generation module 806 is further configured to: determine a target second initial video frame, and determine at least one intermediate pixel point from the target second initial video frame, wherein the target second initial video frame is any one of the initial second video frames; obtain initial pixel motion feature information, and adjust the initial pixel motion feature information according to each intermediate pixel point and the first initial video frame, and obtain target pixel motion feature information corresponding to each intermediate pixel point; and determine motion feature information between the first initial video frame and the target second initial video frame according to each target pixel motion feature information.Optionally, the motion feature generation strategy includes a motion feature adjustment function; the motion feature information generation module 806 is further configured to: process the intermediate pixel points according to the initial pixel motion feature information to obtain a reference pixel point; obtain a pixel difference value corresponding to the reference pixel point according to the motion feature adjustment function, the intermediate pixel points and the reference pixel points; adjust the initial pixel motion feature information according to the pixel difference value to obtain reference pixel motion feature information; use the reference pixel motion feature information as the initial pixel motion feature information, and continue to execute the step of processing the intermediate pixel points according to the initial pixel motion feature information to obtain the reference pixel point, until the adjustment stop condition is met. Optionally, the motion feature adjustment function includes a motion feature area adjustment function or a motion feature overall adjustment function; the motion feature information generation module 806 is further configured to: determine an initial pixel point corresponding to the reference pixel point in the first initial video frame; when the motion feature adjustment function is a motion feature area adjustment function, obtain a pixel difference value based on the reference pixel point and the intermediate pixel point; when the motion feature adjustment function is a motion feature overall adjustment function, obtain an error pixel point corresponding to the intermediate pixel point in each reference second initial video frame, and obtain a pixel difference value based on each error pixel point, the intermediate pixel point and the reference pixel point, wherein the reference second initial video frame is a second initial video frame other than the target second initial video frame. Optionally, the target video generation module 808 is further configured to: determine a target second reference video frame, wherein the target second reference video frame is any one of the second reference video frames; obtain mapped motion feature information, wherein the mapped motion feature information is motion feature information between the first reference video frame and the target second reference video frame; process the target second reference video frame according to the mapped motion feature information to generate a to-be-mixed video frame corresponding to the target second reference video frame; and obtain a target video frame corresponding to the first reference video frame according to a preset video frame mixing rule, the first reference video frame, and the to-be-mixed video frames. Optionally, the target video generation module 808 is further configured to: determine at least one to-be-processed pixel point and a to-be-processed pixel region corresponding to each to-be-processed pixel point in the target second reference video frame, wherein a pixel region includes at least one pixel point; process each to-be-processed pixel region according to the mapped motion feature information to determine a target pixel point corresponding to each to-be-processed pixel point, and determine a to-be-mixed video frame corresponding to the target second reference video frame based on each target pixel point.Optionally, the video frame mixing strategy includes a mixing table query strategy; the target video generation module 808 is further configured to: obtain a target video frame corresponding to the first reference video frame according to a preset video frame mixing rule and the first reference video frame and each video frame to be mixed, including: constructing a mixing table according to each first reference video frame and each video frame to be mixed corresponding to each first reference video frame; based on the mixing table query strategy, querying at least one intermediate mixing result corresponding to the first reference video frame from the mixing table; and mixing each intermediate mixing result to obtain a target video frame corresponding to the first reference video frame. Optionally, the video processing device also includes an interpolation module, which is configured to: obtain at least one initial key video frame in the initial video, and determine a reference key video frame corresponding to the initial key video frame from a reference video corresponding to the initial video; generate key video frame feature information between each initial key video frame based on a preset key video frame feature generation strategy; obtain a transition video frame corresponding to each reference key video frame based on each reference key video frame and the key video frame feature information corresponding to each reference key video frame; and generate a target video based on each reference key video frame and the transition video frame corresponding to each reference key video frame. Optionally, the video frame includes at least one pixel point; the interpolation module is further configured to: determine a first initial key video frame and a second initial key video frame, and determine a reference transition video frame based on the first initial key video frame and the second initial key video frame, wherein the first initial key video frame is any one of the initial key video frames, and the second initial key video frame is a key video frame adjacent to the first initial key video frame; determine an initial transition pixel point in the reference transition video frame; obtain initial pixel transition feature information, determine a first reference pixel point and a second reference pixel point based on the initial pixel transition feature information and the initial transition pixel point, and adjust the initial pixel transition feature information based on the first reference pixel point, the second reference pixel point, and the initial transition pixel point to generate target pixel transition feature information corresponding to each first initial pixel point and the second reference pixel point; and generate key video frame feature information between each initial key video frame based on each target pixel transition feature information.Optionally, the key video frame feature generation strategy includes a transition video frame adjustment function; the interpolation module is further configured to: process the initial transition pixel point based on the initial pixel transition feature information to obtain a first reference pixel point and a second reference pixel point; obtain a transition pixel point difference value corresponding to the reference pixel point based on the transition video frame adjustment function, the initial transition pixel point, the first reference pixel point, and the second reference pixel point; adjust the initial pixel transition feature information based on the transition pixel point difference value to obtain reference pixel transition feature information; use the reference pixel transition feature information as the initial pixel transition feature information, and continue to process the initial transition pixel point based on the initial pixel transition feature information to obtain the first reference pixel point and the second reference pixel point until an adjustment stop condition is met. Optionally, the video processing device further includes a target video sending module configured to: send the target video to a target user. Optionally, the video processing device further includes a target video adjustment module configured to: receive video adjustment information sent by the target user for the target video; adjust the target video according to the video adjustment information to obtain the adjusted target video; and return the adjusted target video to the target user. Optionally, the video processing device further includes a user prompt adjustment module configured to: generate video adjustment prompt information based on the target video; send the target video and the video adjustment prompt information to the target user; receive video adjustment information sent by the target user based on the video adjustment prompt information; adjust the target video according to the video adjustment information to obtain the adjusted target video; and return the adjusted target video to the target user. The above is a schematic diagram of a video processing device according to this embodiment. It should be noted that the technical solution of this video processing device and the technical solution of the aforementioned video processing method share the same concept. For details not described in detail in the technical solution of the video processing device, refer to the description of the technical solution of the aforementioned video processing method. By applying the solution of the embodiments of the present disclosure, a motion feature generation strategy that does not require a deep learning model or a machine learning model is employed. By calculating the motion feature information between frames in the initial video, the computational power consumption during video processing can be reduced while ensuring the smoothness of the generated video.Furthermore, by calculating the motion feature information between frames in the reference video, the motion relationship between frames in the smooth initial video can be determined. This motion feature information can then be used to process the non-smooth reference video, avoiding the problem of video flickering caused by discontinuities between frames. This allows the flickering reference video to be rendered as a smooth target video. Furthermore, a special tree-like blending table stores the blending rules required for each frame. This allows for more efficient video processing, reducing time and computing power consumption and improving processing efficiency. Furthermore, a key video frame feature generation strategy is used to process key video frames in the smooth initial video to obtain features of transition frames between key video frames in the smooth video. This allows key video frame extraction from the non-smooth reference video. Key video frames of the reference video are obtained and processed, and smooth transition frames are inserted between the key video frames of the reference video, allowing the flickering reference video to be rendered as a smoother target video. FIG9 shows a block diagram of a computing device 900 according to one embodiment of the present disclosure. The components of computing device 900 include, but are not limited to, a memory 910 and a processor 920. Processor 920 is connected to memory 910 via a bus 930. A database 950 is used to store data. Computing device 900 also includes an access device 940 that enables computing device 900 to communicate via one or more networks 960. 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 940 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 902.11 wireless local area network (WLAN) wireless interface, a Worldwide Interoperability for Microwave Access (Wi-MAX) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, or a Near Field Communication (NFC) interface. In one embodiment of the present disclosure, the aforementioned components of the computing device 900 and other components not shown in FIG. 9 may also be connected to each other, for example, via a bus. It should be understood that the computing device structure block diagram shown in FIG. 9 is for illustrative purposes only and does not limit the scope of the present disclosure. Those skilled in the art may add or replace other components as needed. The computing device 900 may be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., a tablet computer, a personal digital assistant, a laptop computer, a notebook computer, a netbook, etc.), a mobile phone (e.g., a smartphone), A wearable computing device (e.g., a smartwatch, smart glasses, etc.) or other type of mobile device, or a stationary computing device such as a desktop computer or personal computer (PC). The computing device 900 may also be a mobile or stationary server. The processor 920 is configured to execute the following computer-executable instructions, which, when executed by the processor, implement the steps of the aforementioned video processing method, an automatic question-answering method, or a video processing method applied to a cloud server. The above is a schematic diagram of a computing device in this embodiment. It should be noted that the technical solution of the computing device and the technical solution of the aforementioned video processing method, an automatic question-answering method, or a video processing method applied to a cloud server are based on the same concept. 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 aforementioned video processing method, an automatic question-answering method, or a video processing method applied to a cloud server.One embodiment of the present disclosure further provides a computer-readable storage medium storing computer-executable instructions. When executed by a processor, these computer-executable instructions implement the steps of the aforementioned video processing method, an automatic question-answering method, or a video processing method applied to a cloud server. The above is an exemplary embodiment of a computer-readable storage medium according to this embodiment. It should be noted that the technical solution of this storage medium shares the same concept as the technical solution of the aforementioned video processing method, an automatic question-answering method, or a video processing method applied to a cloud server. 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 aforementioned video processing method, an automatic question-answering method, or a video processing method applied to a cloud server. Another embodiment of the present disclosure further provides a computer program product, including a computer program / instructions. When executed by a processor, these computer program / instructions implement the steps of the aforementioned video processing method, an automatic question-answering method, or a video processing method applied to a cloud server. The above is an exemplary embodiment of a computer program according to this embodiment. It should be noted that the technical solution of this computer program shares the same concept as the technical solution of the aforementioned video processing method, an automatic question-answering method, or a video processing method applied to a cloud server. For details not described in detail in the technical solution of the computer program, please refer to the description of the technical solution of the aforementioned video processing method, an automatic question-answering method, or a video processing method applied to a cloud server. The above description describes specific embodiments of the present disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different from that described in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous. The computer instructions comprise computer program code, which may be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium may include any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a mobile hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electric carrier signal, a telecommunications signal, and a software distribution medium. It should be noted that the content of the computer-readable medium may be appropriately increased or decreased based on the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media do not include electric carrier signals and telecommunications signals.It should be noted that, for ease of description, the aforementioned method embodiments are described as a series of actions. However, those skilled in the art should understand that the embodiments of the present disclosure are not limited by the order of the actions described, as certain steps may be performed in a different order or simultaneously, depending on the embodiments of the present disclosure. Furthermore, those skilled in the art should also understand that the embodiments described in this specification are preferred embodiments, and the actions and modules described are not necessarily required for the embodiments of the present disclosure. In the above embodiments, the descriptions of each embodiment are given with emphasis. For portions not described in detail in a particular embodiment, reference should be made to the relevant descriptions of other embodiments. The preferred embodiments disclosed above are merely intended to help illustrate the present disclosure. The alternative embodiments do not describe all details in detail, nor do they limit the invention to the specific implementations described. Obviously, many modifications and variations are possible based on the content of the embodiments of the present disclosure. The present disclosure selects and describes these embodiments in detail to better explain the principles and practical applications of the embodiments of the present disclosure, thereby enabling those skilled in the art to better understand and utilize the present disclosure. The present disclosure is limited only by the claims and their full scope and equivalents.

Claims

Claims 1. A video processing method, comprising: Acquire a first initial video frame in an initial video, and at least one second initial video frame corresponding to the first initial video frame; Determining a reference video corresponding to the initial video, and determining a first reference video frame corresponding to the first initial video frame and a second reference video frame corresponding to each second initial video frame, wherein the number of video frames of the reference video is the same as the number of video frames of the initial video; generating motion feature information between the first initial video frame and each second initial video frame based on a preset motion feature generation strategy; The first reference video frame is processed according to each motion feature information to obtain a target video frame corresponding to the first reference video frame, and a target video is generated according to at least one target video frame.

2. The method according to claim 1, wherein: A video frame includes at least one pixel point; based on a preset motion feature generation strategy, motion feature information between the first initial video frame and each second initial video frame is generated, including: determining a target second initial video frame, and determining at least one intermediate pixel point from the target second initial video frame, wherein the target second initial video frame is any one of the initial second video frames; obtaining initial pixel motion feature information, and adjusting the initial pixel motion feature information according to each intermediate pixel point and the first initial video frame, and obtaining target pixel motion feature information corresponding to each intermediate pixel point; and determining motion feature information between the first initial video frame and the target second initial video frame according to each target pixel motion feature information.

3. The method of claim 2, wherein: The motion feature generation strategy includes a motion feature adjustment function; adjusting the initial pixel motion feature information based on each intermediate pixel and the first initial video frame includes: processing the intermediate pixel according to the initial pixel motion feature information to obtain a reference pixel; obtaining a pixel difference value corresponding to the reference pixel based on the motion feature adjustment function, the intermediate pixel, and the reference pixel; Adjusting the initial pixel motion feature information according to the pixel point difference value to obtain reference pixel motion feature information; The reference pixel motion feature information is used as initial pixel motion feature information, and the step of processing the intermediate pixel point according to the initial pixel motion feature information to obtain the reference pixel point is continued until an adjustment stop condition is met.

4. The method of claim 3, wherein: The motion feature adjustment function includes a motion feature area adjustment function or a motion feature overall adjustment function; obtaining a pixel difference value corresponding to the reference pixel point according to the motion feature adjustment function, the intermediate pixel point and the reference pixel point, including: when the motion feature adjustment function is a motion feature area adjustment function, obtaining a pixel difference value according to the reference pixel point and the intermediate pixel point; when the motion feature adjustment function is a motion feature overall adjustment function, obtaining an error pixel point corresponding to the intermediate pixel point in each reference second initial video frame, and obtaining a pixel difference value according to each error pixel point, the intermediate pixel point and the reference pixel point, wherein the reference second initial video frame is a second initial video frame other than the target second initial video frame.

5. The method of claim 1, wherein Processing the first reference video frame according to each motion feature information to obtain a target video frame corresponding to the first reference video frame includes: determining a target second reference video frame, wherein the target second reference video frame is any one of the second reference video frames; obtaining mapping motion feature information, wherein the mapping motion feature information is motion feature information between the first reference video frame and the target second reference video frame; processing the target second reference video frame according to the mapping motion feature information to generate a to-be-mixed video frame corresponding to the target second reference video frame; and obtaining a target video frame corresponding to the first reference video frame according to a preset video frame mixing rule and the first reference video frame and each to-be-mixed video frame.

6. The method of claim 5, wherein: Processing the target second reference video frame according to the mapped motion feature information to generate a to-be-mixed video frame corresponding to the target second reference video frame includes: determining at least one to-be-processed pixel point and a to-be-processed pixel region corresponding to each to-be-processed pixel point from the target second reference video frame, wherein the pixel region includes at least one pixel point; Each pixel area to be processed is processed according to the mapped motion feature information, a target pixel corresponding to each pixel to be processed is determined, and a video frame to be mixed corresponding to the target second reference video frame is determined according to each target pixel.

7. The method according to claim 5, wherein: The video frame mixing strategy includes a mixing table query strategy; obtaining a target video frame corresponding to the first reference video frame according to a preset video frame mixing rule and the first reference video frame and each video frame to be mixed, including: constructing a mixing table according to each first reference video frame and each video frame to be mixed corresponding to each first reference video frame; based on the mixing table query strategy, querying at least one intermediate mixing result corresponding to the first reference video frame from the mixing table; and mixing the intermediate mixing results to obtain a target video frame corresponding to the first reference video frame.

8. The method according to claim 1, wherein: Before obtaining a first initial video frame in an initial video and at least one second initial video frame corresponding to the first initial video frame, the method further includes: obtaining at least one initial key video frame in the initial video, and determining a reference key video frame corresponding to the initial key video frame from a reference video corresponding to the initial video; generating key video frame feature information between each initial key video frame based on a preset key video frame feature generation strategy; obtaining a transition video frame corresponding to each reference key video frame based on each reference key video frame and the key video frame feature information corresponding to each reference key video frame; and generating a target video based on each reference key video frame and the transition video frame corresponding to each reference key video frame.

9. The method according to claim 8, wherein A video frame includes at least one pixel point; based on a preset key video frame feature generation strategy, key video frame feature information between each initial key video frame is generated, including: determining a first initial key video frame and a second initial key video frame, and determining a reference transition video frame based on the first initial key video frame and the second initial key video frame, wherein the first initial key video frame is any one of the initial key video frames, and the second initial key video frame is a key video frame adjacent to the first initial key video frame; determining an initial transition pixel point in the reference transition video frame; obtaining initial pixel transition feature information, and obtaining initial pixel transition feature information based on the initial pixel transition feature information and the initial transition image. The method comprises the steps of: determining a first reference pixel point and a second reference pixel point, adjusting the initial pixel transition feature information according to the first reference pixel point, the second reference pixel point, and the initial transition pixel point, generating target pixel transition feature information corresponding to each of the first initial pixel point and the second reference pixel point; and generating key video frame feature information between each initial key video frame according to each target pixel transition feature information.

10. The method of claim 9, wherein: The key video frame feature generation strategy includes a transition video frame adjustment function; determining a first reference pixel and a second reference pixel based on the initial pixel transition feature information and the initial transition pixel point, and adjusting the initial pixel transition feature information according to the first reference pixel point, the second reference pixel point, and the initial transition pixel point to generate target pixel transition feature information corresponding to each of the first initial pixel point and the second reference pixel point, including: processing the initial transition pixel point according to the initial pixel transition feature information to obtain a first reference pixel point and a second reference pixel point; obtaining a transition pixel point difference value corresponding to the reference pixel point according to the transition video frame adjustment function, the initial transition pixel point, the first reference pixel point, and the second reference pixel point; adjusting the initial pixel transition feature information according to the transition pixel point difference value to obtain reference pixel transition feature information; using the reference pixel transition feature information as the initial pixel transition feature information, and continuing to process the initial transition pixel point according to the initial pixel transition feature information to obtain the first reference pixel point and the second reference pixel point until an adjustment stop condition is met.

11. The method according to any one of claims 1 to 10, wherein: After generating the target video, the method further includes: sending the target video to a target user.

12. The method according to claim 11, wherein: After sending the target video to the target user, the method further includes: receiving video adjustment information sent by the target user for the target video; adjusting the target video according to the video adjustment information to obtain an adjusted target video, and returning the adjusted target video to the target user.

13. The method according to any one of claims 1 to 10, wherein: After generating the target video, the method further includes: Based on the target video, video adjustment prompt information is generated; the target video and the video adjustment prompt information are sent to a target user; video adjustment information sent by the target user based on the video adjustment prompt information is received; the target video is adjusted according to the video adjustment information to obtain an adjusted target video, and the adjusted target video is returned to the target user.

14. An automatic question-answering method, comprising: Receive a video modification instruction, wherein the video modification instruction carries an initial video and an initial question text; parse the video modification instruction to determine the initial question text, and generate an initial video modification prompt text based on the initial question text; input the initial video and the initial video modification prompt text into a video modification model to obtain a reference video generated by the video modification model; obtain motion feature information between a target first video frame and each second video frame in the reference video based on the initial video, wherein the target first video frame is any one video frame in the reference video, and the second video frame is at least one video frame corresponding to the target first video frame in the reference video; adjust each target first video frame in the reference video based on the motion feature information corresponding to each target first video frame in the reference video to generate a target video, and generate video modification answer information based on the target video and the initial question text.

15. A video processing method, applied to a cloud server, comprising: A video processing instruction sent by a GAL 1 device at the receiving end, wherein the video processing instruction carries an initial video and a reference video corresponding to the initial video; obtaining a first initial video frame in the initial video, and at least one second initial video frame corresponding to the first initial video frame; obtaining a reference video corresponding to the initial video, and determining a first reference video frame corresponding to the first initial video frame and a second reference video frame corresponding to each second initial video frame, wherein the number of video frames of the reference video is the same as the number of video frames of the initial video; generating motion feature information between the first initial video frame and each second initial video frame based on a preset motion feature generation strategy; processing the first reference video frame according to each motion feature information, and obtaining a first reference video frame corresponding to the first reference video frame. and generating a target video according to at least one target video frame, and returning the target video to the terminal device.

16. A computing device, comprising: memory and processor; The memory is used to store computer programs / instructions, and the processor is used to execute the computer programs / instructions. When the computer programs / instructions are executed by the processor, the steps of the method described in any one of claims 1-13. 14 or 15 are implemented.

17. A computer-readable storage medium storing a computer program / instruction, wherein the computer program / instruction, when executed by a processor, implements the steps of the method according to any one of claims 1 to 13, 14 or 15.

18. A computer program product comprising a computer program / instructions, which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 13, 14 or 15.

Images