Microphone connection interaction display method and device, equipment and medium

Abstract

The invention relates to the field of network live broadcast, and discloses a microphone-connected interactive display method and device, equipment and medium, and the method comprises the steps: displaying digital substitute models corresponding to a plurality of activity participating anchors participating in a dance competition activity on a graphical user interface of a microphone-connected live broadcast room; in response to the dance starting event of the current activity link, corresponding to each activity participation anchor, determining a matched target dance action based on an interaction control event triggered by a supporting user of the activity participation anchor in the current activity link; driving the corresponding digital substitute model of the corresponding activity participating anchor to perform dance according to the target dance action; and in response to the dancing ending event of the current activity link, determining match result data according to control result data triggered by the interaction control event of each activity participating anchor, and displaying the match result data on the graphical user interface. According to the invention, the user experience of interactive competitive activities in the microphone-connected live broadcast room can be comprehensively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of network live broadcast, in particular to a method and device for displaying a live microphone interaction, equipment and a non-transitory computer readable medium. BACKGROUND

[0002] In the live microphone room of the existing network live broadcast platform, interactive competition activities generally use text barrage, gift list or simple animation icons as feedback means, forming a long-term solidified technical path. With the high-frequency online of similar activities, users quickly develop tolerance to the visual stimulation of traditional forms, and the platform needs to continuously increase the value of gifts or the complexity of animations to barely maintain activity. The scissors difference between operating costs and experience benefits is widening. The following analyzes the technical problems exposed by the existing technology from three dimensions of interaction form, visual presentation and emotional closed loop:

[0003] In terms of interaction form, the traditional scheme updates the competition results of the host and the audience in the form of a two-dimensional ranking list, digital increase or decrease, or barrage screen, and the user must continuously translate abstract numbers into cognition of the host's performance. This additional cognitive conversion path significantly prolongs the attention focusing time, leading to viewing fatigue. Since the platform lacks other carriers that can dynamically change with real-time interaction data, the instantaneous flashing of gift special effects and barrage colors can only bring millisecond-level stimulation, and cannot form a continuous plot development. The diminishing marginal utility of the technology is thus revealed.

[0004] In terms of visual presentation, due to the limitation of two-dimensional plane information carrying capacity, traditional gift animations cannot change the spatial level according to the real-time fluctuations of interaction data, resulting in monotonous and repetitive visual stimulation. As the same animation template is repeatedly called, users develop visual tolerance, and the platform is forced to increase the value of gifts or superimpose more complex animation sequences, but it is still difficult to break through the upper limit of emotional peaks that can be provided by two-dimensional planes, and the operating cost rises while the activity improves limitedly.

[0005] In terms of emotional closed loop, the existing technology fails to provide a visual link that can bind the host's identity and the audience's behavior in real time. Users cannot intuitively perceive the immediate impact of their behavior on the host, and the host also lacks a visual medium to feedback the audience's support, so the interaction link is broken, resulting in a thin activity atmosphere, a shortened user stay time, and a long-term low activity curve of the platform.

[0006] In view of the above deficiencies of the traditional technology, the present application attempts to explore a more effective solution. SUMMARY

[0007] The present application aims to solve the above problems and provide a method and device for displaying a live microphone interaction, equipment and a non-transitory computer readable medium.

[0008] According to an aspect of the present application, a method for presenting a live interaction is provided, comprising the following steps:

[0009] displaying a plurality of digital avatar models corresponding to a plurality of activity participating anchors participating in a dance performance competition activity on a graphical user interface of a live broadcast room;

[0010] in response to a dance performance start event of a current activity stage, determining a target dance action corresponding to each activity participating anchor based on an interaction control event triggered by a support user of the activity participating anchor at the current activity stage;

[0011] driving the corresponding digital avatar model of the corresponding activity participating anchor according to the target dance action, so that the digital avatar model performs dance according to the target dance action;

[0012] in response to a dance performance end event of the current activity stage, determining competition result data according to control result data triggered by the interaction control event of each activity participating anchor, and displaying the competition result data on the graphical user interface.

[0013] According to another aspect of the present application, a device for presenting a live interaction is provided, comprising:

[0014] an activity preparation module configured to display a plurality of digital avatar models corresponding to a plurality of activity participating anchors participating in a dance performance competition activity on a graphical user interface of a live broadcast room;

[0015] an action determination module configured to, in response to a dance performance start event of a current activity stage, determine a target dance action corresponding to each activity participating anchor based on an interaction control event triggered by a support user of the activity participating anchor at the current activity stage;

[0016] an action execution module configured to drive the corresponding digital avatar model of the corresponding activity participating anchor according to the target dance action, so that the digital avatar model performs dance according to the target dance action;

[0017] a result display module configured to, in response to a dance performance end event of the current activity stage, determine competition result data according to control result data triggered by the interaction control event of each activity participating anchor, and display the competition result data on the graphical user interface.

[0018] According to another aspect of the present application, a device for presenting a live interaction is provided, comprising a central processing unit and a memory, wherein the central processing unit is configured to call and run a computer program stored in the memory to perform the steps of the method for presenting a live interaction.

[0019] According to another aspect of the present application, a non-volatile readable storage medium is provided, which stores a computer program implemented according to the live-mic interactive display method in the form of computer readable instructions. When the computer program is invoked and run by a computer, the steps included in the method are executed.

[0020] Compared with the prior art, the present application presents a digital avatar model for each active anchor in the live-mic live streaming room in real time, which can switch dance actions according to interactive control events, directly converts abstract interactive data streams into continuous shape performances in three-dimensional space, and makes users no longer need to interpret the gift list or barrage again, so that the attention is continuously locked on the progressive changes of the avatar dance. At the same time, the dance actions of the digital avatar model are mapped with the interactive control events, forming emotional peaks fluctuating with the interactive data, so that the platform can maintain freshness and stimulation without constantly increasing gifts or special effects, and the scissors difference between operation cost and experience benefit can be converged. The digital avatar model has the dual functions of anchor identity visualization and audience feedback medium, each interaction can be used to drive the corresponding digital avatar model to complete the dance steps, forming an emotional closed loop from the audience to the digital avatar model and then to the anchor, significantly improving the atmosphere concentration and user stay time in the live streaming room, and driving the continuous rise of platform activity. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 An exemplary network architecture suitable for applying the live-mic interactive display method of the present application;

[0022] Figure 2 An exemplary live-mic live streaming room graphical user interface of the present application;

[0023] Figure 3 A flowchart of an embodiment of the live-mic interactive display method of the present application;

[0024] Figure 4 A principle block diagram of the live-mic interactive display device of the present application;

[0025] Figure 5 A structural diagram of a live-mic interactive display device used in the present application. DETAILED DESCRIPTION

[0026] To facilitate the understanding of various embodiments of the present application, an exemplary network architecture and application scenario are introduced.

[0027] The network architecture of the present application is designed to support network live streaming service, and to realize real-time transmission of live streaming information by efficiently transmitting various data and live streaming, including image stream and audio stream. For example, Figure 1In the illustrated architecture, the architecture includes a live server 83, a media server 85, a terminal device 90 of a live user, and a terminal device 92 of a spectator user. The live server 83 is responsible for managing and distributing live content, and the media server 85 processes and transmits media data. The live user performs live streaming through the terminal device 90, and the spectator user watches the live streaming through the terminal device 92. These terminal devices run live streaming applications to access the network live streaming service and interact with the servers to use the service. The servers 83 and 85 run corresponding programs to support the normal operation of the live streaming service. The live streaming interaction display method of the present application can be implemented as a computer program product, which is flexibly deployed to run in the terminal device 92 of the spectator user and / or the server according to the function it implements.

[0028] In the exemplary live streaming business scenario of the present application, there are multiple live rooms connected to each other at the same time, and they are displayed in the form of live streaming rooms. Through the coordination of the live server and the media server, multiple audio and video streams are spliced into the same picture and synchronously pushed to all spectator terminals in the live streaming room. The spectator terminals display these pictures in the graphical user interface of the live streaming room, so that the spectator user can watch the live streaming pictures of each host user in the live streaming room, and the spectator user can also interact through the graphical user interface.

[0029] Two or more host users can act as activity participating users and jointly initiate an interactive competition activity in the live streaming room, so that the spectator users supporting the activity participating users become support users of the activity participating users. By triggering an interactive control event in the interactive competition activity, the user activity level of the live streaming room is improved.

[0030] In the exemplary application scenario corresponding to the technical solution provided by the present application, the interactive competition activity in the live streaming room is innovatively implemented as a dance performance competition activity performed by digital avatar models instead of corresponding activity participating hosts. By mapping and matching the interactive control events triggered by the support users of each activity participating host, the target dance action to be performed by the digital avatar model corresponding to each activity participating host is determined, and then the target dance action is used to drive the corresponding digital avatar model to perform dance, so that the activity interaction process in the dance performance competition activity is represented by the digital avatar model performing dance.

[0031] The dance performance competition activity of the present application can include multiple activity links, and each activity link synchronously displays the digital avatar model corresponding to each activity participating anchor in the graphical user interface of the live streaming room. After the activity link starts, the user is supported to accumulate the current active index to the corresponding anchor by triggering the interactive control event, and the background compares the active level threshold in real time with a pre-designed time window as the statistical granularity, and drives the digital avatar model to switch to the matching target dance action. At the end of the activity link, the competition result data is determined and displayed according to the control result data generated by the interactive control event of each anchor, and each activity participating user waits for the next activity link to be started, thereby forming a continuous competition closed loop that evolves immediately with the user support behavior.

[0032] The length of the timing window and the active level threshold used in each activity link of the dance performance competition activity can be pre-set by the platform operator, or dynamically adjusted according to real-time traffic or festival activity needs. Typically, the timing window can be set to any interval between 10 seconds and 2 minutes, and the active level threshold can be divided into 3 to 5 levels in a multiple increasing manner, for example, with a progressive gradient of 50, 100, 200, 500 corresponding to different intensity dance actions, thereby balancing between low latency and high impact.

[0033] The graphical user interface of the live streaming room of the present application can load and display the live video stream of each anchor in the live streaming room, especially the live video stream of the activity participating user. For the dance performance competition activity, a corresponding dance performance area can be opened in the graphical user interface, and the background image can be pre-arranged in the dance performance area, and then the digital avatar model of each activity participating anchor is displayed in the dance performance area. The dance performance area can be arranged near the live video stream playback window, and in specific scenarios where the dance performance area needs to be highlighted, the live video stream playback window can also be hidden or replaced, and only the digital avatar model in the dance performance area can indicate each corresponding activity participating anchor. In some other embodiments, the animation flow of each digital avatar model in the dance performance area can also be processed by the media server to generate a corresponding image flow, replace the original live video stream, and transmit it to the terminal device of the audience user in the live streaming room for playback and display. In this case, the original live video stream playback window is reused as the dance performance area, which can save the system overhead of the terminal device side.

[0034] The digital avatar model of the present application can be generated based on the same three-dimensional model data corresponding to each active user, and different active users can use different three-dimensional model data according to their selected angles. In order to facilitate the driving of the digital avatar model, the present application can correspond to various preset dance actions, and prepare the corresponding bone key frame sequence for each dance action. When the corresponding digital avatar model needs to be driven, the target dance action sequence is called, the digital avatar model changes the pose according to the key point data in each bone key frame, and then the image frame under the pose is rendered, and the dance action process of the digital avatar model is presented by playing frame by frame.

[0035] The interactive control event of the present application includes but is not limited to any one or any combination of the following interactive control events: user speech event, user barrage event, user gift event, user sharing event, and user attention event triggered by the support user in the live streaming room. In the present application, a single interactive control event can be used to control the digital avatar model to switch dance actions, or multiple interactive control events can be used to jointly control the digital avatar model to switch dance actions.

[0036] In the present application, the process of determining the target dance action according to the interactive control event can be implemented on the terminal device side or on the media server side. When implemented on the terminal device side, the terminal device is locally responsible for loading and rendering the digital avatar model; when implemented on the media server side, the media server is responsible for loading and rendering the digital avatar model, and the corresponding animation stream is transmitted to the terminal device, and the terminal device outputs the animation stream to the corresponding performance area for display.

[0037] In order to present the competition results of the performance competition activity in the live streaming room, the present application can determine the competition ranking of each active user based on the control result data generated by one or more interactive control events, and display the competition ranking in the form of visual data, images, animations, etc. to the graphical user interface of the live streaming room.

[0038] Based on the above preparatory explanation of the network architecture and exemplary scenarios of the present application, the following will continue to develop detailed descriptions of various specific embodiments of the present application.

[0039] Please refer to Figure 2 In one embodiment, the live streaming interactive display method of the present application comprises:

[0040] Step S5100, display the digital avatar model corresponding to each active anchor participating in the performance competition activity in the graphical user interface of the live streaming room;

[0041] When the anchor user in the live streaming room clicks or touches the dancing activity control on the graphical user interface or other equivalent operation entry, the terminal device of the anchor user reports an activity participation event to the server. After receiving at least two such events, the server confirms these anchors as activity participation anchors and issues an activity start notification message to all audience user terminal devices in the live streaming room.

[0042] After receiving the notification message, the audience terminal switches the current live streaming room graphical user interface to a dancing competition activity style, which adds a dancing area to the original video window or below, as shown in FIG. 5, for collectively displaying all digital avatar models. The size, background and layout of the dancing area can be determined by a style package issued by the server. The style package can include various data as needed, such as background images, light parameters, camera configuration and safety margin information, etc. After the terminal device side parses the style package, it can render and generate a show stage atmosphere image in the dancing area according to the data, and subsequently set each digital avatar model as the foreground image of the stage atmosphere image, for example, as shown in FIG. 6. Figure 3 Figure 3

[0043] The source of the digital avatar model can have two complementary implementation examples. In the first implementation example, the server undertakes all the calculations: after confirming the activity participation anchors, the server instantiates the digital avatar models corresponding to the activity participation anchors, then starts the rendering pipeline, continuously renders the digital avatar models according to the pre-planned camera track, and outputs the corresponding animation stream to the audience user terminal devices, which only need to decode and present in the dancing area. The advantage of this implementation example is centralized calculation and low terminal load, which is suitable for low-end devices or environments with limited network bandwidth.

[0044] In the second implementation example, the terminal device completes the calculation: after receiving the list of activity participation anchors, the terminal device downloads the basic three-dimensional model and bone data corresponding to each activity participation anchor from the server, and then instantiates the required number of digital avatar models locally. The terminal device directly uses the local GPU to perform bone animation and lighting calculation based on the basic three-dimensional model and bone data, and outputs the rendering result to the dancing area through the OpenGL interface. The advantage of this implementation example is lower latency, as the anchor's facial expression changes can be immediately reflected on the avatar, and it also saves server rendering and streaming bandwidth, which is suitable for high-end devices and local area network scenarios.

[0045] ​​The display state of the dance performance area is dynamically determined by the number of anchors. When there are only two active participating anchors, the dance performance area is divided into left and right, each displaying one digital avatar model; when there are three or more anchors, a surround layout can be used, and the digital avatar models are arranged at the edge of a virtual circular stage, and the camera can be fixed on a track to surround the display, ensuring the visibility of each avatar. If any active participating anchor exits during the activity, the server immediately issues an update instruction to the terminal, the dance performance area is re-arranged, and the remaining avatars automatically fill the vacancy, maintaining the balance of the picture.

[0046] Since the specific activity link of the dance performance competition activity has not yet started, at this time the digital avatar models can be set to a static state, or set to a certain default dance performance state, to distinguish from the subsequent performance effect of driving the digital avatar models according to the target dance action.

[0047] Step S5200, in response to the dance performance start event of the current activity link, for each active participating anchor, based on the interactive control event triggered by the support users of the active participating anchor in the current activity link, determine the matching target dance action;

[0048] The dance performance start event can be triggered and broadcasted by the server. When all active participating anchors have confirmed their positions and the dance performance area rendering is complete, it means that the activity link participation request of each active participating anchor has reached the server, and the server accordingly issues a dance performance start message to all audience terminals in the live streaming room. The dance performance start message can carry the start and end timestamps, link time length, and active gear threshold table of the current activity link, thereby announcing the start of the interactive competition phase. After the audience terminal receives the message, it immediately presents a countdown or a flashing light visual prompt on the graphical user interface to inform the support users that the competition is in progress.

[0049] The interactive control event in the present application includes but is not limited to any one or any combination of the user gift event, the user barrage event, the user follow event, the user share event, and the user tap like event. Its triggering process follows a uniform terminal-server-terminal closed loop. For example, after the audience clicks on the gift icon, inputs a barrage, or performs other interactive operations on the graphical user interface, the terminal encapsulates the operation as an event data packet, which at least contains the event type, event value, unique identifier of the target active participating anchor, and timestamp, and sends it to the server through an encrypted signaling channel; the server verifies the legality of the data packet and immediately returns an acknowledgment message, at the same time, the event is classified into the event queue of the corresponding anchor, and the total number of events of the anchor in the current timing window determined according to the preset statistical granularity can be calculated in real time as needed, as the current active index.

[0050] In order to effectively map the interactive control event to the target dance action, the server internally maintains a dance action index table in which mapping relationship data from the interactive control event directly or indirectly mapped to the matching dance action identifier is established, where the dance action identifier indicates the corresponding dance action and is mapped to the corresponding action data. Whenever any audience user triggers an interactive control event, the server determines the activity participation host supported by the audience user as a support user, and then queries the dance action index table according to the interactive control event to determine the matching dance action as the target dance action to be performed by the digital avatar model of this activity participation host. Subsequently, the action data is downloaded to the audience terminal or media server to drive the digital avatar model to switch the dance posture.

[0051] In an embodiment, the dance performed by the digital avatar model can be preset with various dance styles, such as street dance, national dance, taxi dance, etc. The dance action index table can mark the dance style identifier to which each dance action belongs. In the same activity competition activity or the same activity link, only the target dance action corresponding to the dance style identifier can be selected according to the dance style identifier preset by the activity participation user.

[0052] The action data of the dance action mainly includes a sequence of bone key frames, and can also include rhythm marks for controlling the display rhythm of the bone key frames and transition time length and other related parameters suitable for controlling the digital avatar model as needed. The sequence of bone key frames describes the three-dimensional coordinates and rotation angles of the joints of the digital avatar model at discrete time points, and the bone key frames can be connected by an interpolation algorithm to generate continuous actions. The server pre-stores a plurality of bone key frame sequences corresponding to dance actions, corresponding to different styles and difficulties, such as street dance rotation, mechanical sliding step, and mid-air split, etc. Since the direct or indirect mapping relationship between the sequence of bone key frames and the interactive control event is indicated in the dance action index table before the activity link is started, the server only needs to query to extract, which is fast and efficient.

[0053] In an embodiment, a high-value virtual gift can trigger a target dance action alone, and the virtual gift identifier is directly mapped to a corresponding dance action identifier in a dance action index table. When a support user triggers an interactive control event corresponding to the virtual gift to be given to an active participant user, the corresponding dance action identifier can be quickly determined through the dance action index table, and then the corresponding action data is called according to the dance action identifier to control the corresponding digital avatar model to perform the corresponding dance action. For example, a rocket gift worth 500 diamonds is defined as a special virtual gift. After the server receives the interactive control event corresponding to the gift, the next dance action of the corresponding active participant host is replaced with a preset rocket liftoff dance step. The dance step includes a flame particle special effect and a liftoff displacement trajectory, which is more eye-catching. The action data can also be stored separately from the dance action index table with the gift identifier as the key, so as to realize instant visual feedback of the gift.

[0054] In another embodiment, the server can collect the same type of target interactive control events for each active participant host in units of timing windows, and determine the target dance action according to the collection result. The length of the timing window can be set to be between 10 seconds and 2 minutes, and the upper limit can be no more than the length of the current activity segment. The starting point of the timing window can be when the active participant host receives the target interactive control event. Then, within the timing window, the subsequent target interactive control events are continuously counted. According to the target interactive control events accumulated at the end of the timing window, an activity index is quantitatively determined, and according to the activity index belonging to the activity level, a target dance action that matches it is determined. If the accumulated activity index does not reach the minimum activity level when the timing window arrives, the target dance action can not be determined, that is, the next dance action of the corresponding digital avatar model is not updated. After the timing window ends, a new timing window can be opened as needed according to the above principle for cyclic processing. In this way, the dance action of the digital avatar model can be triggered by updating the dance action according to the same type of interactive control events contributed by multiple support users, and the multi-user interaction performance can be displayed through the digital avatar model.

[0055] In another embodiment, the server can determine the target dance action by combining the two types of interactive control events. To this end, the server can set independent timing windows in parallel for each type of interactive control event of each active participant user based on the previous embodiment, and count them respectively. The weight coefficients are pre-assigned to the activity indexes corresponding to each timing window, and then the two activity indexes obtained by counting at the end of each timing window are weighted and fused into the same activity index, and then the matching target dance action is determined according to the activity index. This way takes into account the uniformity of multiple interactive control events, and effectively guides the interactive activity form of the live streaming room, for example, it can balance the virtual gift economic value and the bullet screen heat of the live streaming room, so that the action switching is closer to the comprehensive popularity.

[0056] Step S5300: Drive the corresponding digital avatar model of the live streamer participating in the activity according to the target dance movement, so that the digital avatar model performs the dance according to the target dance movement;

[0057] After determining the target dance movements, they need to be applied to the corresponding digital stunt double model so that viewers in the live dance broadcast room can see a continuous, smooth, and interactive performance. To achieve this, the skeletal keyframe sequence in the motion data is first treated as the driving information source. Then, the rendering pipeline transforms these discrete 3D pose data into a continuous 2D image frame sequence, and finally outputs it to the graphical user interface frame by frame at a predetermined frame rate. In one embodiment, the skeletal keyframe sequence consists of several timestamps, 3D coordinates of joints, and rotation quaternions. Spherical linear interpolation or spline interpolation algorithms are used to generate intermediate poses between adjacent frames to ensure smooth movement. The interpolation step size can be determined by both the target frame rate and the keyframe interval. For example, at 60fps output, if the interval between two keyframes is 500ms, 29 frames of transition poses need to be inserted in between. After interpolation, the GPU applies joint transformations to the vertex mesh of the digital stunt double model using a skinning algorithm to obtain the stunt double's spatial pose at that moment. Subsequently, a complete image frame is generated through camera transformation, perspective projection, lighting calculation, and texture mapping. Repeat the above process for each frame to obtain the dance image frame sequence of the digital stunt double model.

[0058] At the deployment level, there are two complementary implementation methods: terminal device-driven and server-driven. In the terminal device-driven implementation of the digital stunt double model, the complete motion data of the target dance movement, namely the skeletal keyframe sequence along with rhythm markers and transition durations, is sent by the server to the viewer's terminal and stored in its local cache. When the digital stunt double model needs to be driven to perform the target dance movement, the terminal-side logic directly reads the corresponding motion data in the cache and calls the local GPU pipeline to complete the rendering. The entire process requires no additional network round trips, and the end-to-end latency is less than 50ms. This implementation method is suitable for scenarios that are sensitive to network jitter or have sufficient terminal computing power.

[0059] In the server-driven digital stunt double model implementation, motion data is stored on the server. When a target dance move needs to be executed, the server uses the skeletal keyframe sequence of the target dance move to drive the digital stunt double model to obtain the corresponding image frame sequence and encapsulate it into an animation stream, which is then pushed to the viewer's terminal. The terminal is only responsible for decoding and displaying the animation in the performance area. This approach significantly reduces CPU and GPU usage, making it suitable for low- to mid-range devices or large-scale concurrent scenarios.

[0060] Regarding the action execution strategy, this application also provides several embodiments. In one embodiment, an immediate-effective mode is adopted: whenever a new target dance action is determined, the current dance action of the corresponding digital avatar model is immediately interrupted, and the digital avatar model's posture is seamlessly switched to the new action's starting frame. This embodiment has the fastest response, and users can obtain the corresponding perception more quickly after executing interactive control events, but visual jumps may occur when the action styles differ greatly. Another embodiment introduces an action execution queue: each digital avatar model maintains a first-in-first-out queue. Newly arrived target dance actions are enqueued, and the next action is taken and executed only after the current action finishes playing or reaches a preset interruption point; the queue depth can be configured to 1 to 3 layers, which avoids the tearing feeling caused by frequent switching and ensures the timeliness of interactive feedback. Furthermore, the action execution queue can also support priority queueing, where dance actions corresponding to high-value gifts can be set to high priority and placed at the head of the queue for immediate playback, while actions triggered by ordinary bullet comments are still queued in order.

[0061] In one embodiment, to achieve a smooth transition between dance movements, transition correction can be performed when the current movement is nearing its end. Specifically, the skeletal pose of the last frame of the current movement is first obtained as the source pose, and the skeletal pose of the first frame of the next target dance movement is read as the target pose. The displacement and rotation differences of the corresponding joints are calculated. Then, starting from the source pose, a Bézier curve or cubic spline interpolation trajectory is generated within a preset transition window of, for example, 250ms, so that the joint differences smoothly decay to zero during the transition time. The corrected skeletal keyframe sequence replaces the first few frames of the original target dance movement, thereby ensuring visual continuity. In a further embodiment, if the current dance movement and the target dance movement have a significant difference in style, an additional general transition movement, such as standing adjustment or rotating in place, can be inserted to mask the abruptness of the style switch.

[0062] Step S5400: Respond to the dance end event of the current activity segment, and determine the competition result data to be displayed on the graphical user interface based on the control result data triggered by the interactive control events of each participating anchor.

[0063] When the server determines that the set time for the current activity segment has been reached, it sends a dance performance end message to all terminals in the live stream room. This message carries control result data generated by interactive control events during the segment's duration for each participating streamer, as well as the competition result data generated accordingly. Upon receiving the message, the terminals synchronously display the final competition results in the graphical user interface, thus completing a full dance performance competition loop.

[0064] The generation process of control result data directly corresponds to the type of interactive control event. For user gift-giving events, the server continuously monitors gift arrival records during the event. Whenever a viewer sends a virtual gift to a participating streamer, the server immediately adds the preset support index of that gift to the streamer's cumulative support score, obtaining the total support score. The support index can be either the gift's face value or a weighted coefficient preset by the platform, such as a rocket gift with a weighted coefficient of 10 and a rose gift with a weighted coefficient of 1. At the end of the event, the cumulative support score constitutes the control result data for the participating streamer in the gift dimension. If the platform also collects user bullet screen events, the server counts the number of bullet screen messages or the sentiment score of the bullet screen text during the event, multiplies the statistical results by preset weights to obtain the control result data for the bullet screen dimension. The values ​​of the two dimensions can be linearly weighted or exponentially weighted according to the platform strategy to form a single total support score as a comprehensive indicator. When the event ends, each streamer receives a unique and sortable total support score, which is included in the control result data.

[0065] Based on the control result data of each participating anchor in each activity segment, the corresponding competition result data for that activity segment can be further compared and determined. The competition result data can be generated by the server after sorting the control result data, and may take the form of any one or more of the following: a ranking list, percentage difference, advancement or elimination indicators. The ranking list can be directly mapped to the anchor's avatar and serial number; the percentage difference is obtained by calculating the percentage difference between the control result data of the first-place anchor and the data of the other anchors; and the advancement indicator is presented as a graphic badge. The server can encapsulate the above information into a result display message, the message body of which may include the anchor's unique identifier, ranking, numerical value, badge resource address, and sound effect resource address, among other information determined as needed.

[0066] After the terminal device analyzes the result display message, it can be visualized on the graphical user interface of the live streaming room. For example, a semi-transparent results panel can be overlaid above the dance area. The panel can list the streamer's avatar, ranking number, support value, and percentage difference, etc., as needed. If the streamer wins first place, a golden halo effect plays around the avatar, and a victory sound effect is triggered. If the streamer is eliminated, the avatar turns gray and an elimination badge is overlaid. The panel fades out automatically after five seconds, and the dance area returns to its initial state, waiting for the next activity segment.

[0067] It should be noted that for dance competitions that include multiple stages, in addition to displaying the competition results data for each stage, the server can further integrate the control results data of all stages after all stages have ended to determine the final winner as the corresponding competition result data for the entire dance competition. Following the same principle as above, this result data is then displayed in the graphical user interface of the live streaming room, thus concluding the entire dance competition.

[0068] As can be seen from the above embodiments, this application comprehensively enhances the user experience of interactive competitive activities in live streaming rooms through innovative interactive competitive modes, achieving multiple technical advantages, including but not limited to:

[0069] Firstly, this application introduces digital avatar models corresponding one-to-one with the participating streamers in the live-streaming room, allowing interactive competition to unfold directly in three-dimensional space, no longer limited to numbers and text on a two-dimensional plane. After a dance start event is triggered, the digital avatar model can switch target dance moves based on the interactive control events triggered by the corresponding participating user in the current activity segment, thus immediately transforming abstract interactive data into visible physical performance. Viewers no longer need to interpret leaderboards or bullet comments; their attention is continuously focused on the constantly changing dance moves of the digital avatar, significantly shortening the cognitive conversion path and avoiding viewing fatigue caused by traditional methods.

[0070] Secondly, because digital avatar models can rotate, translate, and change posture in a three-dimensional coordinate system, the intensity and frequency of interactive control events can be directly mapped to the complexity and rhythm of dance movements, forming a coherent and progressively unfolding plot. For example, when support rapidly increases, the digital avatar model can switch to higher-level dance movements, amplifying the emotional peak as data fluctuates; when support levels level off, the rhythm of the movements can be slowed down accordingly to maintain viewing tension. Thus, the platform no longer needs to constantly increase the value of gifts or add special effects; it can continuously provide fresh stimulation through the layered changes in three-dimensional space, effectively narrowing the gap between operating costs and experiential benefits.

[0071] Furthermore, the digital avatar model simultaneously serves as both a visualization of the streamer's identity and a real-time feedback mechanism for viewer behavior. Each viewer interaction can drive the corresponding streamer's digital avatar model to complete specific dance steps, allowing users to intuitively perceive their immediate impact on the streamer. The streamer, in turn, can use the digital avatar model's dance performance to express real-time gratitude to the viewer, forming a real-time emotional loop. This replaces the fragmented interaction links of traditional technologies, significantly enhancing the atmosphere of the live stream and increasing user dwell time, thereby driving a sustained upward trend in the platform's activity curve.

[0072] Based on any embodiment of the method in this application, digital avatar models corresponding to each of the multiple live streamers participating in the dance competition are displayed in the graphical user interface of the live streaming room, including:

[0073] Step S5110: When the number of anchors who trigger the event in the live streaming room reaches the target number, the graphical user interface of the live streaming room is switched to the dance competition event style, which includes a dance area.

[0074] The server monitors in real-time the activity participation events triggered by each host user in the live-streaming room, covering all host terminals with established connection relationships within the live-streaming room. Activity participation events are generated when a host clicks a dance activity control on the terminal's graphical user interface. The event data packet contains at least a unique host identifier, a timestamp, and an event type identifier. Upon receiving the event data packet, the server writes it to the activity participation event queue in memory and updates the counter using the unique host identifier as the key. When the counter reaches a preset target number, the server determines that the dance competition activity can start. The target number can be set as needed by the platform's operational strategy or flexibly set by any host in the live-streaming room; a typical value is two to eight hosts, or it can be directly set to the total number of hosts in the current live-streaming room. After the counter meets the condition, the server constructs a notification message. The message body may contain relevant information that can be accessed by terminal devices as needed, such as an activity start flag, a host list, a dance area style index, and the duration of each segment. This message is pushed to all terminal devices in the live-streaming room via a long connection channel or room broadcast protocol.

[0075] After receiving the notification message, the terminal device retrieves the corresponding layout data based on the dance area sample index and switches the current graphical user interface to the dance competition activity style. In one exemplary application of the dance competition activity style, the original video stream window is first paused from its top-mounted state, and then the dance area is inserted into the layout tree in full-screen or picture-in-picture format. The size, background, lighting parameters, and camera trajectory of the dance area are determined by the style index issued by the server. The style index points to a pre-set style package on the platform, which is cached locally on the terminal as a binary resource or downloaded on demand. If the corresponding style package already exists on the terminal, it is parsed directly; otherwise, the terminal first pulls the style package via CDN before parsing and loading it. After parsing, the rendering thread generates the initial image of the dance area based on the stage size and light texture recorded in the style package, and simultaneously starts the fade-in playback of background music, completing the interface switch.

[0076] In scenarios where the number of live streamers changes dynamically, the server continuously monitors for exit events. If a live streamer actively exits the live stream during the countdown or in-process phase, the server immediately removes that live streamer from the event participation queue, decrements the counter by one, and generates an update notification that is pushed to all terminals. After receiving the update notification, the terminals reread the latest live streamer list and rearrange the performance area. For example, if there were originally four live streamers arranged in a grid pattern, after one exits, the remaining three are automatically adjusted to a horizontally evenly distributed layout, and the camera trajectory is updated synchronously to ensure image balance and visual continuity.

[0077] Step S5120: Instantiate the digital stand-in model corresponding to each participating anchor, and update the facial mesh texture of the corresponding digital stand-in model of the participating anchor with the facial image corresponding to the participating anchor.

[0078] The instantiation of the digital avatar model is determined by the scene calculation mode. In server-driven mode, the server retrieves the corresponding 3D model data from the model repository based on the list of participating broadcasters, instantiates it in memory, and then streams the instantiated digital avatar model and motion data to the terminal; the terminal is only responsible for decoding and rendering. In terminal-driven mode, after receiving the broadcaster list, the terminal first checks if the corresponding model file exists in the local cache directory. If it does not exist, it pulls it through CDN, then calls the graphics interface to instantiate the model in GPU memory, and immediately begins skeletal binding and skinning calculations. The switching between the two modes is determined by the platform dynamically issuing configuration bits based on device performance and network quality, ensuring that low-end devices use server streaming and high-end devices use local instantiation.

[0079] Digital avatar models can use unified 3D model data or be configured with independent 3D model data for each participating streamer. Under the unified model approach, all streamers share the same skeletal hierarchy and mesh topology, with differences only achieved by replacing texture maps and scaling parameters; this approach significantly reduces package size and is suitable for rapid deployment. Under the independent model approach, each streamer has their own unique skeleton and mesh, allowing for the preservation of personalized body shapes or clothing accessories; the platform maintains a unique entry for each streamer in the model repository, indexed by a unique streamer identifier.

[0080] The live streamer participating in the event can modify their digital avatar model using their own facial images to better represent their own image. The facial images can be preset by the live streamer or captured in real-time from their live video stream. In a preset implementation, the live streamer can upload one or more facial images, typically their own image, such as three standard photos (front, left, and right sides). The server compresses these images into a 256×256 texture and binds it to the live streamer's account, directly retrieving the texture when needed. In a real-time capture implementation, the terminal device or server continuously extracts facial images from the live streamer's video stream: first, a face detection algorithm is used to locate key facial features; then, based on the current head pose of the digital avatar model, the frame closest to the pose angle is selected as the target facial image. If the pose changes drastically, a transition texture is generated by interpolation between adjacent frames to avoid abrupt changes. The two schemes can be dynamically switched: preset textures are activated when network bandwidth is insufficient (e.g., below a preset idle bandwidth threshold), and real-time texture updates are activated when bandwidth is sufficient (e.g., above the idle bandwidth threshold).

[0081] The process of updating the facial mesh texture of a digital avatar model using facial images can be completed on the GPU. In one embodiment, the facial region of the digital avatar model is pre-divided into four meshes: forehead, eyes, nose, and mouth. Each mesh stores UV coordinate indices. During the update, the facial image is cropped into four regions to generate four sub-textures. Then, the graphics interface is called to write the sub-textures into the texture cache corresponding to the UV coordinates. After writing is complete, the rendering pipeline is notified to rebind the updated textures. If a real-time update method is used, the change in the head pose of the digital avatar model can be detected frame by frame. A new round of texture replacement is triggered only when the change angle exceeds a threshold, thereby reducing the GPU load. After the update is completed, the digital avatar model presents a face consistent with the broadcaster.

[0082] Throughout the lifespan of the digital avatar model, the above process can be used to associate the digital avatar model with the facial image of the participating anchor, so that the digital avatar model always displays the designated facial image of the participating anchor.

[0083] Step S5130: Start the rendering thread of the digital stunt double model and continuously render the digital stunt double model with updated facial mesh texture to the dance area for display.

[0084] After the digital stunt double model is instantiated, the rendering thread can be started. The rendering thread then displays the dance area layout and the digital stunt double model in the graphical user interface. The thread first creates an independent rendering context, binds it to the frame buffer object corresponding to the dance area, and then enters a continuous loop: each frame, it reads the digital stunt double model with updated facial textures from the model cache, performs skeletal animation calculations, and passes the joint transformations driven by the skeletal keyframes to the vertex shader via the matrix colorist to complete vertex skinning; next, the fragment shader samples the updated facial and clothing textures, applies the corresponding lighting parameters for the dance area, and outputs the final pixel color. The rendering result is written to the frame buffer and then presented to the surface of the dance area via the swap chain. The frame rate can be kept synchronized with the display refresh, typically 60 FPS. If running in server-driven mode, the rendering thread resides on a cloud GPU node, and the generated image frame sequence is encoded using H.264 or H.265 and then streamed to the terminal, which only decodes and displays it. If running in terminal-driven mode, the rendering thread executes directly on the local GPU, skipping the network encoding step, reducing latency to a single frame time. The thread lifecycle can be synchronized with the activity phase. After the activity ends, the rendering context is destroyed, and the resources are recycled to the object pool, waiting for the next phase to reinitialize.

[0085] The above embodiments achieve a one-to-one correspondence between the streamer's physical identity and the virtual avatar by seamlessly mapping the real-time facial images of the participating streamer to the facial mesh texture of the digital avatar model. This allows viewers to directly see the avatar's dance synchronized with the streamer's real face within the 3D dance area, thus eliminating the sense of separation between identity and appearance found in traditional 2D gift lists or bullet comments. This mapping process is completed in the GPU with frame-level precision, ensuring the instantaneous consistency of texture angle and lighting when the head posture changes, avoiding visual misalignment. The server-terminal collaborative instantiation strategy further ensures low-latency presentation, allowing viewers to receive immediate and intuitive feedback on the avatar's face and dance moves with every interaction. This forms a seamless emotional loop from user behavior to streamer's face to dance performance, continuously capturing attention and significantly improving the immersion and activity of the live streaming room.

[0086] Based on any embodiment of the method in this application, updating the facial mesh texture of the corresponding digital avatar model of the participating anchor with the facial image corresponding to the participating anchor includes:

[0087] Step S5121: Extract facial images corresponding to each angle from the live video stream of the participating anchors.

[0088] After receiving the live video stream from the participating anchor, the server or terminal initiates the face detection and pose estimation module. It scans each image frame in the live video stream, using a keypoint detection model to output the 3D coordinates of the eyes, nose tip, and corners of the mouth. Based on this, it calculates the pitch, yaw, and roll angles of the head orientation and divides the pose space into multiple discrete angle slots with a 15-degree step size. When the anchor's head pose falls into a certain angle slot, that image frame is extracted as the face image at the corresponding angle, while the pose angle is retained as metadata. The extracted face image is immediately sent to the GPU-side texture cache, establishing a circular buffer based on angle indexes. Each slot retains only the latest frame, ensuring real-time and unique textures. If network bandwidth is limited, the image is compressed using YUV420 and downsampled to 256×256; if bandwidth is sufficient, the original resolution is retained to preserve detail.

[0089] In one embodiment, this step can be automatically triggered after each dance competition event and before the start of its event segment. After obtaining a facial image frame in each angle slot, the recognition of the live video stream is paused to save resource overhead. In other words, as long as multiple facial images from the live video stream of the corresponding event participants are obtained before entering the event segment, they can serve the subsequent event segments to cover the entire dance competition event, avoiding excessive consumption of runtime resources due to frequent face detection of the live video stream.

[0090] This approach applies to both the unified model and the independent model scenarios described above: in the unified model, all anchors share the same set of angle slot queues; in the independent model, an independent queue is maintained for each anchor. Ultimately, this multi-angle facial image set provides source textures that perfectly match the real-time pose for subsequent texture replacement, achieving high-fidelity facial reconstruction of the digital avatar model.

[0091] Step S5122: Monitor the real-time head posture of the digital avatar model of the live streamer participating in the event when it performs the target dance action, and determine the face image whose angle is closest to the real-time head posture as the target face texture image.

[0092] When the digital avatar model performs various movements, such as dance moves, within the performance area, its internal skeletal keyframe sequence drives continuous rotation and displacement of the head joints. By reading the transformation matrix corresponding to the head joints in the skeletal keyframes in real time, the pitch, yaw, and roll angles of the head in the world coordinate system for the current frame are calculated, and the three-axis angles are combined into the current head pose vector. This vector is then matched with the nearest neighbor of the circular buffer index established in the previous step. First, the yaw angle is compared; if the difference is less than 15 degrees, the facial image texture in the corresponding slot is directly selected. If the yaw angle difference exceeds a threshold, the pitch and roll angles are compared until the slot with the smallest sum of squared angle differences is found, and the corresponding facial image texture is determined.

[0093] After selecting the desired face texture, the GPU binds the face texture in the selected slot to the corresponding UV region of the face mesh texture, completing the determination of the target face texture image. If no matching precise angle is available in the buffer, spherical linear interpolation can be used to generate a transition texture between two adjacent frames as the target face texture image, ensuring that there are no jumps when the head moves rapidly.

[0094] Real-time pose monitoring can be triggered periodically by the rendering frame rate. In terminal-driven mode, the calculation is completed by the local GPU, while in server-driven mode, the pose vector is sent back by the cloud GPU and then the texture index is sent down. Both methods ensure that the digital avatar model's face and the anchor's real head pose are synchronized at the sub-second level.

[0095] Step S5123: Replace the target face texture image with the face mesh texture of the digital avatar model to update the face mesh texture of the digital avatar model.

[0096] Once the target face texture image is determined, the replacement process begins. A uniformly sized texture cache is reserved on the GPU for the face mesh, and this cache is divided into texel regions corresponding one-to-one with UV coordinates. The replacement operation is performed by copying the target face texture image row by row to the corresponding UV regions of this cache: first, the graphics interface is called to lock the cache; then, texture pixel values ​​are written row by row; after writing is complete, the cache is unlocked, and the rendering pipeline is notified to rebind the updated texture cache. If the texture size is inconsistent with the original UV size of the mesh, the GPU performs bilinear resampling during writing, scaling the 256×256 or original resolution texture to the required resolution of the mesh, ensuring edge alignment and no jagged edges.

[0097] For real-time update scenarios, updates are triggered periodically on rendering frames. A replacement is performed when the head pose change exceeds a set angle threshold; if the pose is stable, the write is skipped to reduce GPU load. In a unified model scenario, all streamers share the same texture cache, and different streamers are distinguished by offsetting the UV start address during replacement. In an independent model scenario, each streamer has an independent texture cache, and replacement only applies to the current streamer's cache. After replacement, the digital avatar model's facial area immediately displays the latest texture, ensuring that the facial image seen by the audience in the performance area is synchronized with the streamer's current real face, with no visible delay or tearing.

[0098] The above embodiments pre-capture and cache multi-angle facial images from the live video stream. Only one calculation and storage are needed before the dance performance begins. Throughout the entire activity cycle, the texture that best matches the current head posture can be selected in real time using nearest neighbor matching of posture vectors and angle slots. This eliminates the redundant overhead of re-parsed the video stream in each rendering frame and ensures that the face of the digital avatar model always matches the latest appearance of the streamer. On the GPU side, textures are written on demand using single-cache, dual-cache, or independent caching strategies, and updates are triggered by posture thresholds. This avoids unnecessary VRAM operations and minimizes the synchronous consumption of computing resources, bandwidth, and rendering load. Thus, while maintaining high-fidelity visual synchronization, the overall system resource consumption is significantly reduced.

[0099] Based on any embodiment of the method in this application, and based on the interactive control events triggered by the supporting user of the live streamer during the current activity segment, a matching target dance movement is determined, including:

[0100] Step S5210: In the current activity phase, using a preset time window as the statistical granularity, accumulate the total number of interactive control events triggered by all supported users for each participating anchor as the current activity indicator;

[0101] The mechanism for creating and maintaining timed windows operates independently in server memory, with each participating broadcaster as the granularity. Whenever the server receives the first target interactive control event for a participating broadcaster, it immediately records the window start timestamp for that broadcaster and begins accumulating the number of events as a current activity metric.

[0102] The timing window length can be preset by the platform, with typical values ​​being 10 seconds, 30 seconds, or 60 seconds, and the upper limit not exceeding the remaining duration of the current activity segment. During the duration of the timing window, the server continuously monitors the target interaction control events generated by the same participating broadcasters. Each received event increments the current activity metric through atomic operations, ensuring concurrency safety. If the window is full before the end of the activity segment, the server immediately closes the timing window and fixes the current activity metric; if the segment ends early, any unfulfilled timing windows are also forcibly closed to prevent data idling. After the timing window closes, the server reclaims the memory space for reuse in the next round of the activity.

[0103] In scenarios that support multiple types of interactive control events, the server maintains an independent timing window for each type of event, including but not limited to gift event window, bullet screen event window, like event window, etc. These can be set as needed and run in parallel. Each event accumulates its total event count, and at the end of the corresponding window, its total event count is multiplied by the corresponding preset weight coefficient for weighting. The weighted result is then used as the current activity indicator.

[0104] Step S5220: Compare the current activity index with the preset activity level threshold to determine the target activity level that matches the current activity index;

[0105] Once the current activity metric is fixed at the end of the timeout window, the level determination process can begin. The server maintains an activity level threshold table in memory, which is a set of integer or floating-point boundary values ​​arranged in ascending order. Typically, it can be configured with three, five, or seven levels, such as a progressive gradient of 50, 100, 200, and 500. Each boundary value corresponds to a target activity level identifier, which is a consecutive integer or enumerated value, and the order is mapped one-to-one with the dance action identifiers in the dance action index table. Accordingly, the current activity metric is compared with the boundary values ​​in the threshold table one by one until the first boundary value greater than or equal to the current value is found, and the corresponding level is the target activity level. If the current activity metric is lower than the lowest boundary, the default level or an empty level can be returned, indicating that no action switching is triggered.

[0106] When the platform selects to count multiple types of interactive control events in parallel, the server obtains the weighted activity index of each type of event after the timer window is closed. Then, the weighted results are accumulated to form a comprehensive current activity index, and a unified level determination is performed again.

[0107] To adapt to the varying intensity of different activity stages, the activity tier threshold table supports dynamic adjustment. In one embodiment, the platform's backend issues a new threshold array before the start of each stage, and the server atomically replaces it in the memory-mapped area. Subsequent timing windows directly use the new table without requiring a service restart. In another embodiment, the threshold table automatically scales based on the real-time number of online viewers: by monitoring the total number of viewers in the current live stream, it is mapped to a preset scaling factor, and then a fixed threshold is multiplied by the factor to obtain the real-time threshold, achieving a flexible strategy where the higher the number of viewers, the higher the tier. A third embodiment enables a tiered increase during festivals or sporting events, with the threshold rising by a certain percentage after fixed time intervals to continuously boost viewer engagement.

[0108] Step S5230: Based on the target activity level, query the pre-stored dance movement index table to determine the target dance movement that maps to the activity level.

[0109] After determining the target activity level, the server immediately uses that level value as the index key to query the dance movement index table to determine the corresponding dance movement identifier for the target dance movement. The dance movement identifier returned by the query uniquely points to a sequence of skeletal keyframes, which has been pre-generated offline, serialized, and stored in the model repository. Based on the dance movement identifier of the target dance movement, the corresponding skeletal keyframe sequence is retrieved from the model repository to construct motion data, which can then be used to drive the corresponding digital avatar model to perform the dance movement.

[0110] The above embodiments, through the independent creation, accumulation, and closing mechanism of the timing window, allow the server to discretize the continuously influx of interactive control events into measurable activity indicators. These indicators are then mapped to dance action identifiers in one go via an activity level threshold table, ensuring a precise correspondence between interaction intensity and dance complexity. Level determination and action invocation are completed in milliseconds, and the digital avatar model upgrades its dance moves in a step-by-step manner according to the activity indicators. Viewers see a progressive physical narrative driven by their own behavior in real time in the dance area, keeping their attention continuously engaged. The parallel maintenance and weight fusion of the timing window further ensure that multiple types of events contribute synchronously, while dynamic thresholds and step-by-step increases allow the competitive intensity to adaptively increase with the scene. The entire activity can maintain peak emotions without additional special effects or gifts, thus narrowing the gap between operating costs and experience benefits. The platform completes a closed-loop mapping from data to plot with minimal computational overhead.

[0111] Based on any embodiment of the method in this application, in the current activity phase, using a preset time window as the statistical granularity, the total number of interactive control events triggered by all supported users for each participating anchor is accumulated as the current activity indicator, including:

[0112] Step S6100: Respond to the target interaction control event triggered by the user, determine the target virtual gift and the target activity participant streamer specified by the target interaction control event, and deduct the target virtual gift from the target activity participant streamer. The target interaction control event is a user gift-giving event.

[0113] In this embodiment, the user gifting event is used as a target interactive control event to statistically measure the activity level and support of participating users. When a viewer, acting as a supporter, clicks the gift icon of the target virtual gift in the graphical user interface of the live broadcast room and confirms payment, the terminal device immediately generates a user gifting event. The event data packet may contain the unique identifier of the corresponding target virtual gift, the quantity, the unique identifier of the giver, and the unique identifier of the target live broadcaster, and is sent to the server via an encrypted signaling channel.

[0114] When the server receives the data packet, it first queries the gift configuration table using the unique identifier of the target virtual gift to obtain the gift value, supported indicator weights, and special effects resource index. Then, it checks whether the giver's account balance or token balance is sufficient. If it is insufficient, it returns a failure response and terminates the process. If the balance is sufficient, it deducts the gift value from the giver's account and atomically increments the gift income field of the target activity's participating streamer, completing the formal transfer of ownership of the virtual gift.

[0115] Step S6200: In the graphical user interface of the live streaming room, play the gift effects of the target virtual gift;

[0116] Once the server confirms that the target virtual gift has been successfully deducted from the target livestreamer, it immediately sends a gift effect playback command to all viewer terminals in the livestream room. This command may include the effect resource index, the unique identifier of the target livestreamer, the gift quantity, and the duration. Upon receiving the command, the terminal device locates and retrieves the corresponding gift effect playback data from its local cache or storage network path based on the effect resource index. This includes, but is not limited to, particle effects, animation textures, and sound effects data required for the target virtual gift's animation effects. Subsequently, the playback process is initiated in the dance area of ​​the graphical user interface or in the gift overlay of the livestream room.

[0117] In one embodiment, the gift effects of the target virtual gift are superimposed above the dance area. The particle emitter position is bound to the central skeleton of the digital avatar model of the target livestreamer. The emission direction and speed are driven by a three-dimensional vector field in the gift configuration table. Color gradients, scaling animations, and gravity decay are performed during the particle's lifecycle. In another embodiment, the gift effects cover the entire graphical user interface in the form of a full-screen animation. The brightness of the background video stream is briefly reduced, and the animation texture is synthesized frame-by-frame with the background through alpha blending. Simultaneously, stereo sound effects are triggered, and the sound effect volume is proportional to the number of gifts.

[0118] Step S630: Accumulate the support index corresponding to the target virtual gift into the total support index of the target activity participating anchor set in the current activity segment as control result data;

[0119] After confirming that the target virtual gift has been credited to the target livestreamer participating in the activity, the server can add the support metric corresponding to the target virtual gift to the total support score field of the target livestreamer in the current activity phase. The support metric can be understood as the number of votes corresponding to the target virtual gift, and the total support score can be understood as the total number of votes.

[0120] Support metrics are directly provided by the preset values ​​of the corresponding target virtual gifts in the gift configuration table, or calculated by multiplying the gift's face value by a weighting coefficient. For example, a rocket gift has a weight of 10, and a rose gift has a weight of 1. The server maintains a total support variable in memory for each participating streamer. Using the activity phase as the statistical period, the total support metric is obtained by summing up the support metrics of all target virtual gifts received during the current activity phase. After accumulation, this total support metric becomes the control result data, used for subsequent competition result calculations, leaderboard ranking, and reward distribution.

[0121] If multi-dimensional statistics are enabled, that is, the total support is calculated based on various types of target interactive control events, the total support of the gift dimension can be weighted and integrated with the total support of dimensions such as bullet comments and likes according to the corresponding preset weight coefficients to determine a comprehensive total support as the control result data of the corresponding target activity anchor.

[0122] Step S6400: Increment the total number of the first events corresponding to the timer window to which the user's gift-giving event belongs as the current active indicator.

[0123] After the server completes the gift deduction and effect triggering, it injects the user's gift-giving event into the timing window of the corresponding target streamer. If the streamer's timing window does not yet exist, the server creates a new window starting from the current timestamp and maintains an atomic integer variable in memory as the first event total for that window. If the window already exists, it directly performs an atomic increment operation on the variable to ensure numerical consistency in concurrent scenarios. After each increment, the server returns the updated first event total as the current activity metric to the activity tier determination module for comparison with the activity tier threshold in the next step. When the timing window closes, the first event total is fixed and cleared to zero, the window memory is reclaimed to the object pool, and the next timing window starts accumulating from zero again.

[0124] The above embodiments map user gift-giving events to two parallel calculation paths: total support and total first event. This achieves real-time accumulation of gift value and updates to activity metrics at the timer window level within the same data stream. Atomic operations and object pool reuse mechanisms ensure numerical consistency and efficient memory reclamation in high-concurrency scenarios, avoiding lock contention and memory leaks. Operations such as gift deduction, special effects playback, support accumulation, and activity metric refresh can be completed sequentially in milliseconds, providing immediate feedback to viewers through both visual gift effects and quantifiable tier progression, significantly shortening the cognitive journey. Total support is used directly as control result data for subsequent rankings and rewards, while the total first event serves as the current activity metric driving dance move upgrades. These two metrics complement each other and do not duplicate calculations, achieving a closed-loop transformation from interactive behavior to competitive scenarios without additional server resources. This reduces operating costs while continuously amplifying user engagement and platform activity.

[0125] Based on any embodiment of the method in this application, in the current activity phase, using a preset time window as the statistical granularity, the total number of interactive control events triggered by all supported users for each participating anchor is accumulated as the current activity indicator, including:

[0126] Step S7100: Respond to the target interactive control event triggered by the user, and determine the target activity participant and bullet screen text pointed to by the target interactive control event, wherein the target interactive control event is a user bullet screen event;

[0127] In this embodiment, user bullet screen events are used as target interactive control events to statistically measure the activity of users participating in the activity. When a viewer submits bullet screen content through the input control in the live broadcast room and confirms the sending, the terminal device generates a user bullet screen event. The event data packet includes the bullet screen text, the sender's unique identifier, the target activity participant's unique identifier, and a timestamp, and is uploaded to the server via an encrypted signaling channel.

[0128] After receiving the data packet, the server first confirms whether the target event's participating streamer is currently in the event. If the streamer has left or not joined, the packet is discarded. Once confirmed, the message text is sent to a sensitive word filter for prohibited content detection. If a violation is detected, it is replaced with a default character or rejected outright. The filtered message text and the unique identifier of the target event's participating streamer are written into the message event queue in memory, along with additional metadata such as the sender's level, badge, and fan club identifier, for subsequent style rendering and weight calculation. At this point, the user's message event is officially captured and associated with the corresponding target event's participating streamer, providing a unique and complete data source for subsequent style display, ambient effect triggering, and the accumulation of the second event's total.

[0129] Step S7200: In the graphical user interface of the live streaming room, the bullet screen text is displayed using the personalized style of the team corresponding to the host participating in the target activity.

[0130] Once the server confirms that the bullet screen text has passed the sensitive word filtering, it immediately sends a style rendering instruction to all viewer terminals in the live broadcast room. The instruction can carry the unique identifier of the target live streamer, the bullet screen text content, and additional metadata.

[0131] Upon receiving the instruction, the terminal retrieves the corresponding style record from the locally pre-set personalized style table of the participating streamer based on the unique identifier of the target activity. The record may contain fields such as font family, font size, color value, stroke width, background color block, animation parameters, and badge icon path. The rendering thread writes the bullet screen text into a separate layer according to the style parameters: the text first generates glyph textures using FreeType or the system font engine, then overlays strokes and shadows; the background color block is filled with rectangular primitives and maintains a fixed margin with the text; the badge icon is instantiated and drawn using the GPU, positioned immediately to the left of the text, and its size scales proportionally with the font size. After the layer is generated, a bullet screen scrolling channel is inserted. The channel speed is calculated by dividing the screen width by the visible duration, ensuring a uniform speed as it scrolls below the dance area.

[0132] If 3D perspective is enabled on the terminal, a rotation matrix is ​​added to the bullet screen layer during the vertex shader stage, causing the text to fly in along a curved path. In server-driven mode, layer compositing can be completed in the cloud and pushed as a video stream, which the terminal directly decodes and displays. Style sheets support dynamic updates; the platform can distribute new style packages before the start of an activity, and the terminal atomically replaces the cache, completing a global switch without restarting the application.

[0133] By applying personalized samples of their respective teams to the streamers participating in the target event, the bullet screen effects of streamers participating in different events can be differentiated, making it easier for viewers to quickly distinguish between them.

[0134] Step S7300: In the graphical user interface of the live streaming room, display the atmospheric effects corresponding to the barrage event in the digital stand-in model of the target activity's participating anchor.

[0135] After confirming that the bullet screen text has passed the sensitive word filtering, the server immediately sends an atmosphere effect command to all viewer terminals. This command can carry the unique identifier of the target streamer, the effect resource index, and the trigger intensity field. The terminal device locates the pre-set or real-time downloaded particle system data based on the resource index, and then binds the particle emitter to the head skeleton node of the streamer's digital avatar model, causing the effect coordinates to change synchronously with the real-time head movement. The particle system is instantiated on the GPU and performs color gradients, scaling animations, and radial diffusion throughout its lifecycle. The intensity field directly maps the emission rate to the maximum number of particles, presenting corresponding atmosphere effects through the display of appropriate particle effects, achieving a positive correlation between bullet screen popularity and visual impact.

[0136] In one embodiment, the particle texture uses a PNG sequence pre-multiplied by alpha, sampled cyclically by a fragment shader; in another embodiment, the particles are generated entirely by the shader program, using a noise function to drive displacement and color, reducing video memory usage. In server-driven mode, the particle rendering results are encoded into the animation stream and pushed to the terminal; in terminal-driven mode, the particle system is computed in real time on the local GPU, with latency less than a single frame time. After the effects playback is complete, the terminal automatically reclaims particle instances and releases video memory, ensuring that resources are cleared as playback begins.

[0137] It should be noted that the ambient effect command in this step can be combined with the rendering command in the previous step into a single bullet screen display command and broadcast to each terminal device in the live broadcast room. After the terminal device parses and obtains the corresponding data, it can execute the display of bullet screen text and ambient effects based on this data.

[0138] Step S7400: Increment the total number of the second events corresponding to the timer window to which the user's bullet screen event belongs as the current active indicator.

[0139] While driving the live stream to play the corresponding bullet screen text and ambient effects for user bullet screen events, the server injects these events into the second event total counter for the target streamer. If the streamer's timing window does not yet exist, the server creates a new window starting from the current timestamp and maintains an atomic integer variable in memory as the second event total. If the window already exists, the server directly performs an atomic increment operation on this variable to ensure concurrent write safety. After each accumulation, the updated second event total is immediately sent to the activity level determination module as the current activity indicator and compared with the activity level threshold table to determine the dance move upgrade. When the timing window reaches its preset length or the activity ends, the server fixes the second event total and clears the counter. The memory occupied by the window is reclaimed to the object pool, and the next timing window starts accumulating from zero again.

[0140] The above embodiments achieve low-latency triple synchronization by simultaneously using user bullet screen events for bullet screen text rendering, atmospheric effect triggering, and atomic accumulation of the total number of second events, completing content presentation, visual feedback, and activity metric updates within the same data path; sensitive word filtering and style differentiation mechanisms ensure content security and brand recognition; particle effects bound to skeletal nodes allow bullet screen popularity to be directly converted into dynamic light effects around the avatar model, enhancing the interactive immersion; the timer window is independently maintained and the total number of events is accumulated atomically, ensuring data accuracy under high concurrency and automatic memory reclamation as the process progresses, allowing bullet screen frequency to be mapped to the rhythm of dance moves in real time without additional server resources, continuously amplifying audience participation and significantly improving platform activity.

[0141] Based on any embodiment of the method in this application, the corresponding digital avatar model of the live streamer participating in the corresponding activity is driven according to the target dance movement, so that the digital avatar model performs the dance according to the target dance movement, including:

[0142] Step S5310: Obtain the skeletal keyframe sequence corresponding to the target dance movement;

[0143] After determining the target dance movement, the server locates the corresponding skeletal keyframe sequence from the model repository based on the dance movement identifier. This record can be implemented as a continuous three-dimensional pose data. Each keyframe consists of a time offset, a joint index array, a translation vector, and a quaternion rotation. It has been generated offline through motion capture or manual animation and serialized in binary format and stored in the model repository.

[0144] There are three implementation methods for the acquisition path: In the server-led implementation, the server encapsulates the complete skeletal keyframe sequence into the gear level query result and sends it to the terminal along with a push message; in the terminal-led implementation, the terminal queries the local cache directory for the corresponding pre-set skeletal keyframe sequence file based on the dance motion identifier, and if it is missing, it is incrementally retrieved via CDN; in the hybrid implementation, the skeletal keyframe sequence corresponding to high-frequency gear levels can reside permanently on the terminal, while low-frequency gear levels are sent by the server on demand in sequence. When loading, the terminal device first checks the local cache, and only performs network retrieval if no match is found. After loading, the skeletal keyframe sequence is sent to the animation pipeline. The time offset is used for subsequent interpolation calculations, and the translation vector and quaternion rotation are used to drive the joint transformations of the digital stunt double model, completing the conversion from dance motion identifiers to usable motion data.

[0145] Step S5320: Generate a stunt dance image frame sequence based on the corresponding digital stunt model driven by the skeletal keyframe sequence.

[0146] After the skeletal keyframe sequence enters the animation pipeline, it is first sent to the interpolation module. This module reads the time offset and generates continuous poses between any two frames: if the keyframe interval is greater than the rendering frame interval, spherical linear interpolation is used to calculate joint rotations, linear interpolation is used to calculate translations, and a frame-by-frame transformation matrix is ​​output; if the keyframe interval is less than the rendering frame interval, the pose of the previous keyframe is directly reused to avoid over-computation. The interpolated transformation matrix is ​​written to a unified buffer for use in subsequent skinning stages.

[0147] The skinning stage is executed in parallel on the GPU. The vertex data of the digital stunt double model contains a quadruple of vertex coordinates, normals, weights, and bone indices. Each vertex is bound to a maximum of four bones. The vertex shader reads the transformation matrix buffer, retrieves the corresponding matrix according to the bone index, multiplies it with the weights, and accumulates the results to obtain the final transformation matrix. This matrix is ​​then applied to the vertex coordinates and normals to complete the pose transformation. The fragment shader then samples the face texture, clothing texture, and light map, outputs color and depth information, and generates a complete stunt double dance image frame. The successively generated stunt double dance image frames constitute the corresponding stunt double dance image frame sequence.

[0148] The rendering context employs a double-buffering mechanism: a front buffer for display and a back buffer for drawing. After each frame is rendered, the front and back buffers are atomically swapped via a swap chain, and the viewer's terminal presents the new frame when the vertical synchronization signal arrives, avoiding tearing. In server-driven mode, the rendering process can be completed on the cloud GPU. The generated sequence of stunt double dance image frames is encoded using H.264 or H.265 and pushed to the terminal via a real-time streaming protocol. The terminal only decodes and displays the image. In terminal-driven mode, rendering is completed on the local GPU, skipping encoding and network round trips, keeping end-to-end latency within one frame.

[0149] To smoothly connect dance moves, transition segments can be inserted during action changes. The default transition segment length is 250ms, but it can be dynamically determined by the gift value or the density of comments. The transition algorithm reads the skeletal poses of the last frame of the previous action and the first frame of the next action, calculates the difference in joint displacement and rotation, uses cubic spline interpolation to generate an intermediate pose, and replaces several keyframes before the next action to ensure visual continuity. If the styles of the two actions are too different, a general transition action—such as a spin in place or a standing adjustment—is inserted first before the target action is executed to avoid abruptness.

[0150] Step S5330: Play the sequence of images of the stunt double dance frame by frame in a loop to show the dance performance process of the corresponding digital stunt double model performing the target dance movements.

[0151] After obtaining image frames from the sequence of images representing the stunt double's dance moves, the rendering pipeline enters a continuous playback phase to present the dance performance. The playback loop is driven by the main thread or a separate rendering thread, reading image frames one by one at fixed time steps. The time step is consistent with the device refresh rate: 16.67ms for a 60Hz device and 8.33ms for a 120Hz device. At the beginning of each frame, the next frame is retrieved from the frame queue. If the number of remaining frames in the queue is below a threshold, a preloading thread is triggered to request subsequent frames from the model repository or cache in advance to prevent stuttering.

[0152] After the frame data arrives, the rendering thread writes the image frame to the back buffer, and then calls the exchange link to atomically swap the back buffer with the front buffer. When the vertical synchronization signal arrives, the contents of the front buffer are sent to the display controller, and the audience can then see the continuous dance footage in the dance area. If running in server-dominated mode, the terminal decoder decodes frame by frame according to the RTP timestamp order and sends them to the frame queue. The decoding thread and the rendering thread exchange data through a lock-free circular buffer. If running in terminal-dominated mode, the GPU directly generates frame data, and the frame queue depth is set to three frames to absorb GPU scheduling jitter.

[0153] To eliminate rhythm drift, the difference between the current playback time and the frame timestamp is calculated based on the audio clock or system monotonic clock. If the difference exceeds a threshold, a frame dropping or interpolation strategy is initiated: if a frame arrives early, the rendering thread sleeps until the correct time; if a frame arrives late, the intermediate frame is skipped or a transition frame is inserted to maintain overall rhythm synchronization with the music beat. The rhythm synchronization algorithm times with 1ms precision within the rendering thread, with an error controlled within ±5ms.

[0154] When a dance move changes, the rendering thread checks if the current frame is the first frame of the transition segment. If so, it locks the transition segment length and simultaneously reads the skeletal poses of the last frame of the previous move and the first frame of the next move during the transition. Intermediate frames are generated in real-time using cubic spline interpolation, replacing the first N frames in the queue. N is calculated by dividing the transition length by the frame interval. The frame rate remains constant during the transition, resulting in a smooth visual transition for the viewer. If the transition span is large, a general transition frame sequence is inserted first, followed by the target move frame sequence. The general transition frames are also generated through interpolation, with a fixed length of 250ms to ensure that stylistic differences are masked.

[0155] The end condition for playback is determined by the duration of the activity segment or the duration of the action. When a segment ends, the rendering thread receives a termination signal, stops reading new frames, clears the frame queue, releases texture buffers and swap chain resources, and the dance area returns to its default static screen, waiting for the next round of activities to begin.

[0156] The above embodiments utilize a digital stunt double model driven by skeletal keyframe sequences to complete a closed-loop process of interpolation, skinning, rendering, playback, and rhythm synchronization within a unified animation pipeline. This process converts discrete posture data into a continuous high-frame-rate image sequence in real time with sub-millisecond time precision, and completely eliminates screen tearing through double buffering and vertical synchronization. The server-led, terminal-led, and hybrid acquisition paths balance bandwidth, computing power, and latency as needed, ensuring smooth presentation on devices ranging from low-end to high-end. Interpolation and transition algorithms automatically connect movements with different styles, resulting in seamless transitions and keeping the audience's attention focused on the dance area. Frame queue depth and preloading mechanisms work together to absorb network or GPU jitter, while the beat synchronization algorithm controls the dance rhythm and music beat error within ±5 milliseconds, achieving zero-perceptible latency. Resources are immediately recycled at the end of each segment, minimizing memory and video memory usage. The platform can maintain a highly immersive competitive experience with low computational costs without additional special effects or gift bonuses, significantly reducing operating costs and continuously increasing user dwell time and platform activity.

[0157] Please see Figure 4According to one aspect of this application, a live-streaming interactive display device includes an activity preparation module 5100, an action determination module 5200, an action execution module 5300, and a result display module 5400. The activity preparation module 5100 is configured to display digital avatar models corresponding to multiple participating anchors in a dance competition on the graphical user interface of the live-streaming room. The action determination module 5200 is configured to respond to a dance start event in the current activity segment, and for each participating anchor, determine a matching target dance action based on the interactive control event triggered by the supporting user of that anchor in the current activity segment. The action execution module 5300 is configured to drive the corresponding digital avatar model of the participating anchor according to the target dance action, causing the digital avatar model to perform the dance according to the target dance action. The result display module 5400 is configured to respond to a dance end event in the current activity segment, and determine the competition result data to be displayed on the graphical user interface based on the control result data triggered by the interactive control events of each participating anchor.

[0158] Based on any embodiment of the device in this application, the digital avatar models corresponding to multiple participating anchors in a dance competition are displayed in the graphical user interface of the live streaming room. This includes: an interface switching module, configured to switch the graphical user interface of the live streaming room to a dance competition style when the number of anchors triggering the participation event in the live streaming room reaches a target number, the dance competition style including a dance area; a face compositing module, configured to instantiate the digital avatar model corresponding to each participating anchor and update the facial mesh texture of the corresponding digital avatar model of the participating anchor with the facial image corresponding to the participating anchor; and a rendering startup module, configured to start the rendering thread of the digital avatar model and continuously render the digital avatar model with updated facial mesh texture to the dance area for display.

[0159] Based on any embodiment of the device in this application, the face synthesis module includes: an image extraction module, configured to extract face images corresponding to multiple different angles from the live video stream of the participating anchor; a pose matching module, configured to monitor the real-time head pose of the digital avatar model of the participating anchor when performing a target dance movement, and determine the face image whose angle is closest to the real-time head pose as the target face texture image; and a texture replacement module, configured to replace the target face texture image into the face mesh texture of the digital avatar model, so as to update the face mesh texture of the digital avatar model.

[0160] Based on any embodiment of the device in this application, the action determination module 5200 includes: an indicator statistics module, configured to accumulate the total number of interactive control events triggered by all supported users for each participating anchor in the current activity phase, using a preset time window as the statistical granularity, as the current activity indicator; a level determination module, configured to compare the current activity indicator with a preset activity level threshold to determine a target activity level that matches the current activity indicator; and an action matching module, configured to query a pre-stored dance action index table based on the target activity level to determine a target dance action that maps to the activity level.

[0161] Based on any embodiment of the device in this application, the indicator statistics module includes: a gift-giving response module, configured to respond to a target interactive control event triggered by a user, determine the target virtual gift and the target activity participant streamer specified by the target interactive control event, and deduct the target virtual gift from the target activity participant streamer, wherein the target interactive control event is a user gift-giving event; a gift display module, configured to play the gift effects of the target virtual gift in the graphical user interface of the live streaming room; a control quantification module, configured to accumulate the support indicators corresponding to the target virtual gift into the total support of the target activity participant streamer set in the corresponding current activity segment as control result data; and a gift quantification module, configured to automatically increment and accumulate the total first event corresponding to the timer window to which the user gift-giving event belongs as the current active indicator.

[0162] Based on any embodiment of the device in this application, the indicator statistics module includes: a bullet screen response module, configured to respond to a target interactive control event triggered by a user, and determine the target participating anchor and bullet screen text pointed to by the target interactive control event, wherein the target interactive control event is a user bullet screen event; a bullet screen display module, configured to display the bullet screen text in the graphical user interface of the live streaming room using the personalized style corresponding to the target participating anchor; an atmosphere display module, configured to display the atmosphere effects corresponding to the bullet screen event in the graphical user interface of the live streaming room, corresponding to the digital avatar model of the target participating anchor; and a bullet screen quantification module, configured to automatically increment and accumulate the total amount of a second event corresponding to the timer window to which the user bullet screen event belongs as the current active indicator.

[0163] Based on any embodiment of the device in this application, the result display module 5400 includes: a sequence acquisition module, configured to acquire the skeletal keyframe sequence corresponding to the target dance movement; a drive conversion module, configured to drive the corresponding digital stunt double model to generate a stunt double dance image frame sequence according to the skeletal keyframe sequence; and a playback display module, configured to play the stunt double dance image frame sequence frame by frame in a loop to display the dance performance process of the corresponding digital stunt double model performing the target dance movement.

[0164] Another embodiment of this application also provides a live-streaming interactive display device. For example... Figure 5 The diagram shows the internal structure of an interactive live-streaming display device. This device includes a processor, a computer-readable storage medium, a memory, and a network interface connected via a system bus. The computer-readable, non-volatile storage medium stores an operating system, a database, and computer-readable instructions. The database stores information sequences, and when executed by the processor, these computer-readable instructions enable the processor to implement an interactive live-streaming display method.

[0165] The processor of the interactive live-streaming display device provides computing and control capabilities to support the operation of the entire device. The device's memory can store computer-readable instructions, which, when executed by the processor, cause the processor to perform the interactive live-streaming display method of this application. The device's network interface is used for communication with a terminal.

[0166] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the interactive display device to which the present application is applied. The specific interactive display device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0167] In this embodiment, the processor is used to execute... Figure 4 The specific functions of each module are described, and the memory stores the program code and various data required to execute the above modules or sub-modules. The network interface is used to realize data transmission between user terminals or servers. The non-volatile readable storage medium in this embodiment stores the program code and data required to execute all modules in the interactive display device of this application, and the server can call the server's program code and data to execute the functions of all modules.

[0168] This application also provides a non-volatile readable storage medium storing computer-readable instructions, which, when executed by one or more processors, cause the one or more processors to perform the steps of the interactive live-streaming display method of any embodiment of this application.

[0169] This application also provides a computer program product, including a computer program / instructions that, when executed by one or more processors, implement the steps of the method described in any embodiment of this application.

[0170] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. This computer program can be stored in a non-volatile readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The aforementioned storage medium can be a computer-readable storage medium such as a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM).

Claims

1. A method for interactive live streaming display, characterized in that, include: The digital avatars of the various live streamers participating in the dance competition will be displayed on the graphical user interface of the live stream room. In response to the dance start event of the current activity segment, for each participating streamer, based on the interactive control events triggered by the supporting users of that participating streamer in the current activity segment, determine the matching target dance moves; The corresponding digital avatar model of the live streamer is driven according to the target dance movement, so that the digital avatar model performs the dance according to the target dance movement. In response to the end of the dance performance in the current activity segment, the competition result data is determined and displayed on the graphical user interface based on the control result data triggered by the interactive control events of each participating anchor.

2. The interactive display method according to claim 1, characterized in that, The digital avatars of the various participants in the dance competition will be displayed on the graphical user interface of the live stream, including: When the number of streamers who trigger the event in the live-streaming room reaches the target number, the graphical user interface of the live-streaming room will be switched to the dance competition event style, which includes a dance area. Instantiate digital stand-in models corresponding to each participating anchor, and update the facial mesh texture of the corresponding digital stand-in model of the participating anchor with the facial image corresponding to the participating anchor. The rendering thread of the digital stunt double model is started, and the digital stunt double model with updated facial mesh texture is continuously rendered to the dance area for display.

3. The interactive display method according to claim 2, characterized in that, Updating the facial mesh texture of the corresponding digital avatar model of the livestreamer participating in the event with the facial image of the livestreamer participating in the event includes: From the live video stream of the host participating in the event, facial images corresponding to each angle are extracted from multiple different angles; Monitor the real-time head posture of the digital avatar model of the live streamer participating in the event when it performs the target dance moves, and determine the face image whose angle is closest to the real-time head posture as the target face texture image; The target facial texture image is replaced in the facial mesh texture of the digital avatar model to update the facial mesh texture of the digital avatar model.

4. The interactive display method according to claim 1, characterized in that, Based on the interactive control events triggered by the supporting users of the participating streamer in the current activity segment, the matching target dance moves are determined, including: In the current activity phase, the total number of interactive control events triggered by all supported users for each participating anchor is accumulated using the preset time window as the statistical granularity and used as the current activity indicator. The current activity index is compared with a preset activity level threshold to determine the target activity level that matches the current activity index. Based on the target activity level, query the pre-stored dance movement index table to determine the target dance movement that corresponds to that activity level.

5. The interactive display method according to claim 4, characterized in that, In the current activity phase, using a preset timer window as the statistical granularity, the total number of interactive control events triggered by all supported users for each participating streamer is accumulated as the current activity metric, including: In response to a user-triggered target interaction control event, the system determines the target virtual gift and the target activity participant streamer specified by the target interaction control event, and deducts the target virtual gift from the target activity participant streamer. The target interaction control event is a user gift-giving event. In the graphical user interface of the live streaming room, the gift effects of the target virtual gift are played. The support metrics corresponding to the target virtual gifts are accumulated into the total support level of the target activity participating anchors set in the current activity segment, and used as control result data; The total number of events corresponding to the first event in the timer window to which the user's gift-giving event belongs is incremented and used as the current activity indicator.

6. The interactive display method according to claim 4, characterized in that, In the current activity phase, using a preset timer window as the statistical granularity, the total number of interactive control events triggered by all supported users for each participating streamer is accumulated as the current activity metric, including: Responding to a target interactive control event triggered by a user, the system determines the target activity participant and the bullet screen text to which the target interactive control event points, wherein the target interactive control event is a user bullet screen event; In the graphical user interface of the live streaming room, the bullet screen text is displayed using the personalized style of the team corresponding to the host participating in the target activity. In the graphical user interface of the live streaming room, the digital avatar model of the host participating in the target activity displays the atmospheric effects corresponding to the barrage event; The total number of the second event corresponding to the timer window to which the user's bullet comment event belongs is incremented and used as the current activity indicator.

7. The interactive display method according to any one of claims 1 to 6, characterized in that, Based on the target dance movement, a corresponding digital avatar model of the livestreaming host is driven to perform the dance according to the target dance movement, including: Obtain the skeletal keyframe sequence corresponding to the target dance movement; The skeletal keyframe sequence drives the corresponding digital stunt model to generate a stunt dance image frame sequence. The sequence of images of the stunt double's dance is played in a loop, frame by frame, to demonstrate the dance performance process of the corresponding digital stunt double model performing the target dance movements.

8. A live-streaming interactive display device, characterized in that, include: The event preparation module is set up to display the digital avatar models of the multiple live streamers participating in the dance competition on the graphical user interface of the live stream room. The action determination module is set to respond to the dance start event of the current activity segment. For each participating anchor, based on the interactive control events triggered by the supporting users of the participating anchor in the current activity segment, the module determines the matching target dance action. The motion execution module is configured to drive the corresponding digital avatar model of the live streamer participating in the activity according to the target dance motion, so that the digital avatar model performs the dance according to the target dance motion. The results display module is configured to respond to the dance end event of the current activity segment, and determine the competition results data to be displayed on the graphical user interface based on the control result data triggered by the interactive control events of each participating anchor.

9. An interactive live-streaming display device, comprising a central processing unit and a memory, characterized in that, The central processing unit is used to invoke and run a computer program stored in the memory to perform the steps of the method as described in any one of claims 1 to 7.

10. A non-volatile readable storage medium, characterized in that, It stores, in the form of computer-readable instructions, a computer program implemented according to any one of claims 1 to 7, which, when invoked by a computer, executes the steps included in the corresponding method.

Images