This is a listing of made in China open-source systems and models for creating and operating “digital humans” ranging from speech-driven talking heads and portrait animation to more complex half-body and 3D avatar generation, with an emphasis on making a still image or short video clip speak naturally with accurate lip-sync and controllable expressions, gestures, or body motion. Some entries are primarily models for generating animation, while others are end-to-end toolchains or “panel” style management layers aimed at simplifying local installation, model orchestration, and media pipelines such as audio processing, video synthesis, and voice cloning. Several are positioned for real-time or near-real-time interactive use, sometimes emphasizing lightweight deployment on modest hardware or edge/mobile devices, privacy-preserving on-device operation, and cross-platform integration, while others focus on longer-form video generation, higher resolution output, or more realistic 3D reconstruction and rendering for presenter-style content and interactive assistants in domains like marketing, education, training, and short-form media production.
[51x Jan 2026]
ACTalker: An ICCV 2025 method that makes realistic talking-head videos controlled by audio and/or facial motion.
AIGCPanel: A one-stop local “panel” style system for running digital-human workflows such as video synthesis plus voice synthesis or voice cloning, with simplified model management and installation.
AvatarArtist: Turns one stylized image into an animated 4D avatar that preserves identity and style.
Awesome Digital Human Live2D: wan-h’s open-source Live2D digital human framework integrating ASR, LLM, and TTS.
ChatAnyone: Generates real-time stylized portrait video from one image and audio with face–body sync.
ComfyUI_Sonic: Animates a still face to sing or speak in sync with audio inside ComfyUI.
DiffPortrait360: Generates a realistic, consistent 360° rotatable head (including the back) from a single portrait without retraining or 3D scans.
DisentTalk: Improves cross-lingual talking faces with better lip-sync using spatiotemporal diffusion.
DreamActor-M1: Animates a single photo into a realistic identity-preserving video driven by multi-signal motion guidance, including optional multilingual lip-sync.
DualTalk: Enables lifelike 3D talking-head dialogues with seamless role switching and natural reactions.
Duix.Avatar: An open-source digital-human avatar project positioned for local deployment and API-style use, focused on rapidly generating talking-head style videos from short inputs.
EchoMimic: An open-source audio-driven portrait/upper-body animation project that generates a talking digital human from reference media.
EchoMimic V2: A later iteration of EchoMimic described as improving half-body digital-human animation quality and control from reference inputs.
EdgePersona: A lightweight, fully local digital-human system concept intended to run on ordinary hardware with privacy-preserving on-device operation.
FantasyTalking: Turns one image plus audio into an identity-consistent talking video with natural motion and lip-sync.
Fay: An open-source “talking avatar” style project commonly grouped with digital-human toolkits for generating speech-driven facial animation.
FlowAct-R1: A real-time interactive virtual digital-human technique framed around responsive interaction rather than pre-rendered performance.
GaussianIP: Builds identity-preserving 3D humans from text and a reference image via fast two-stage refinement.
GenHuman: An open-source digital-human project framed around deployable creation of presenter-style digital humans for content production.
GPT-SoVITS: A multilingual voice-cloning and TTS tool that can train from about one minute of audio.
GUAVA: Turns one photo into a detailed, animatable 3D upper-body avatar in ~0.1s, no tuning.
HeyGem: An open-source (or partly open-source, depending on distribution) digital-human pipeline discussed as enabling fast cloning and high-resolution talking-head video synthesis on modest hardware.
HRAvatar: Creates lifelike animatable 3D head avatars from regular video using Gaussian splatting.
HumanDiT: Uses diffusion transformers to turn one image into long, high-quality human motion videos with strong detail.
Hunyuan3D-2: Tencent’s open-source model for generating high-resolution 3D avatars from text or images with consistent textures and easy pipeline integration.
HunyuanCustom: Keeps subjects consistent in personalized videos from text, images, audio, or video.
HunyuanVideo-Avatar: Turns images and audio into multi-character talking videos with emotion.
ImaginTalk: Generates natural speech from silent face video, preserving identity and emotion.
InfiniteTalk: An open-source framework for generating longer-form digital-human videos via a sparse-frame “video dubbing” style approach.
InstantCharacter: Uses diffusion transformers plus an adapter for consistent custom character images.
JoyGen: Uses a two-stage model to generate high-fidelity talking-face video with tight audio lip-sync.
LAM: An open-source project described as generating highly realistic 3D digital humans quickly from limited inputs (often presented as single-image driven).
LatentSync: ByteDance’s open-source diffusion-based lip-sync tool that maps audio to video directly.
LHM: An open-source model described as driving or generating 3D digital humans from a photo, aimed at rapid 3D avatar creation.
Live Avatar: An open-source real-time, audio-driven digital-human generation system described as producing talking digital humans interactively.
LivePortrait: An open-source portrait animation project that makes a still face image “come alive” with speech-driven motion and expression.
MNN-TaoAvatar: An open-source 3D digital-human application built around an on-device inference stack, presented as a practical implementation of TaoAvatar-style methods.
OmniHuman-1: Generates lifelike full-body human videos from one image plus motion signals like audio or video.
OmniSync: Uses diffusion transformers to sync lips to audio across occlusions, poses, and styles.
OpenAvatarChat: An open-source modular “conversational digital human” project combining real-time dialogue with avatar presentation and multimodal interaction.
ReHiFace-S: Guiji AI’s open-source, real-time face-swap model that needs no prior training.
RGBAvatar: Reconstructs photoreal 3D head avatars in real time using compact Gaussian blendshapes.
SEGA: Builds a realistic 3D animated head from one photo, preserving identity and expressions.
SkyReels-V3: A multi-modal video-generation stack that bundles three concrete capabilities placing “virtual character” generation inside a single architecture.
Sonic: An open-source speech-driven digital-human synthesis project described as producing strong lip-sync and expressive talking-head results relative to other open models.
SoulX-FlashTalk: An open-source real-time digital-human talking system described as prioritizing fast startup and smooth, high-FPS conversational video output.
TexTalker: Generates 3D talking avatars from speech with synchronized facial motion and dynamic textures.
THGS: Converts a short single-camera video into a full-body 3D Gaussian-splat avatar with expressive motion.
UniRig: Automates rigging and skinning for diverse 3D models using a large 14k-model Rig-XL dataset.
WeClone: An open-source project that fine-tunes an LLM on WeChat logs to create a personal chat avatar and can also clone the user’s voice from WeChat audio.
Zero-1-to-A: Uses video diffusion to build an animatable head avatar from one photo without tuning.
100 Best Chinese Language Digital Human Projects on GitHub:
[88x Feb 2026]
ArtificialZeng/Fay-digital-human-explained: The Fay Assistant Edition is an important branch of the open-source Fay project, focused on building an open-source solution for intelligent digital assistants; it provides a flexible modular design so developers can customize and combine functional modules, including emotion analysis and NLP ...
Caladog/HeyGem: Contribute to Caladog/HeyGem development by creating an account on GitHub.
ConnectSpace/YodoAI: Build a real-time data-warehouse engine for AI Digital Humans; data comes from Web3 protocol rules and public social-media platforms. https://yodoai.com/ - ConnectSpace/YodoAI.
CrisHY1995/HeadNeRFonDigitalHuman: Although the virtuals built with MetaHuman show ultra-photorealistic imaging and rendering, when we deliberately scrutinize the composites we can always find a few subtly hard-to-describe places that are not quite real; see the image below ...
DivineConcerto/Metaman: An AI that uses the Unity engine\n. 2023-06-02 revised idea: using the Flask framework it can be deployed on a server, receive audio files sent from Unity, convert them to text, submit them to GPT for a reply, and then send the reply back ...
Filifun/digital-wavlipgfp: This project is used to synthesize third-party short videos and can partially substitute for Heygen (because HG review has become increasingly strict). Input source video + audio; output high-definition video with lip-sync. Implementation principle: use Wav2Lip to align audio ...
Fucloud233/sysu-digital-man: This project is developed on top of the open-source Fay project; it modifies and optimizes the LLM part, refactors the logic for calling different LLMs, and introduces a vector database, enabling a fully automated 24-hour online introduction for Sun Yat-sen University ...
Goulandis/MetaHuman: Learning. Contribute to Goulandis/MetaHuman development by creating an account on GitHub.
HansonJames: The general system is an intelligent interactive platform based on deep learning and WebRTC, integrating Azure Avatar rendering, speech recognition and synthesis, and natural language processing; it supports real-time dialogue, knowledge Q&A, and emotion ...
Henry-23/VideoChat: Real-time voice interaction, supporting end-to-end (MLLM-THG) and cascaded (ASR-LLM-TTS-THG) approaches; customizable avatar and voice, supports voice cloning, and first-packet latency as low as 3 seconds. Technical intro: QubitAI post. Simplified Chinese | ...
HuaweiCloudDeveloper/dify-tools: Interactive intelligent Q&A solution, based on Huawei Cloud digital-content production line MetaStudio, the ModelArts Studio “large-model as a service” platform, and Dify for rapid deployment of interactive services, yuez · Dify DeepSeek ...
HumanAIGC-Engineering/OpenAvatarChat: Open Avatar Chat is a modular interactive chat implementation that can run full functionality on a single PC; it currently supports MiniCPM-o as the multimodal language model, or replacing it with a cloud API to implement conventional ASR + LLM + TTS ...
HumanAIGC-Engineering/gradio-webrtc: End-rendering type currently only supports 'gs'. avatar_ws_route, '', end-rendering websocket path. avatar_assets_path, '', end-rendering model asset path. Installation. gradio cc ...
Ikaros-521/AI-Vtuber: Luna AI’s appearance is built from Live2D, VTube Studio, xuniren, and UE5 combined with Audio2Face, EasyAIVtuber, and video players (Easy-Wav2Lip, SadTalker, GeneFace++, MuseTalk, AniTalker ...
Ikaros-521/digital_human_video_player: Luoxi video player with an HTTP API; integrates via the Gradio API with Easy-Wav2Lip, SadTalker, GeneFacePlusPlus, and MuseTalk, and can also be used to play local videos ...
Kedreamix/PaddleAvatar: PaddleAvatar is a generation tool based on the PaddlePaddle deep-learning framework; using multiple Paddle toolkits, it can synthesize your digital images, audio, and video into a realistic video; in addition, ...
LKZMuZiLi/human: Developed using the URP rendering pipeline; supports publishing to all platforms; developed with Unity 2022.3.31 and works out of the box ... Fay open-source framework: https://github.com/xszyou/fay. Large LLMs still require development ...
MicroEngine/Fay_Sales: Fay is a complete open-source project, including the Fay controller and models; it can be flexibly combined into different application scenarios: virtual streamer, on-site sales, product guide, voice assistant, remote voice assistant, interaction, interview ...
MingZheGe/account-digitalPlayer-Unity3D-: Account-opening process with full guided assistance. Contribute to MingZheGe/account-digitalPlayer-Unity3D- development by creating an account on GitHub.
Rayman96/avatar_gpus: Supports concurrent Heygem inference on multi-GPU machines. This is a Flask-based video/audio inference service that can receive video URLs and audio URLs, download the content, run Heygem inference on the best GPU, and return the processed video.
WeHome007/NextCAS-SrvAPI: NextHuman 2D/3D open platform API. Contribute to WeHome007/NextCAS-SrvAPI development by creating an account on GitHub.
WeHome007/NextCAS-UE: NextHuman 3D ultra-realistic/cartoon/heteromorphic UE SDK. Contribute to WeHome007/NextCAS-UE development by creating an account on GitHub.
WeThinkIn/AIGC-Interview-Book: Build a multimodal-AI Su Dongpo, hands-on link, — ; WeClone: a one-stop solution for creating a digital double from chat records, hands-on link · link ; LightX2V four-step distillation model: a 20× faster high-quality video generation revolution ...
YUANZHUO-BNU/metahuman_overview: Main technical整理 ... currently mainly includes avatar/appearance, voice, and dialogue capabilities. The main interaction mode is direct conversation. The following is collected and summarized from multiple aspects to provide a quick-start reference.
YZNYEE/avator: A one-stop virtual live-streaming system. Contribute to YZNYEE/avator development by creating an account on GitHub.
ZEGOCLOUD/digital-human-quick-start-example: This directory contains complete sample code for two application scenarios, helping developers quickly understand the architecture design and implementation for different business scenarios. Sample directories. 1. Interactive chat example (digital-human ...
ZEGOCLOUD/digital_human_paas_quick_start: A complete ZEGO PaaS client quick-start sample project supporting Android, iOS, Web, and Server (Go) platforms, providing full functionality including rendering, multiple driving methods, and task management.
ZEGOCLOUD/zego_aigc_paas_demo: ... The PaaS interface provides the ability to generate tokens. The frontend service is in the src directory, responsible for displaying pages and handling UI interaction logic; after getting the token from the server it calls the PaaS interface to create and use text-driven ...
aidayang/Nova-Virtual-Anchor: This was originally a 2023 project; seeing there is still demand, I redid it. I removed all other functions; the current version only keeps the feature of driving lip-sync speech from an audio file.
aliyun/alibabacloud-avatar-android-demo: The return data of the virtual open platform StartInstance API can be obtained by opening the StartInstance API debug page in the OpenAPI portal, clicking the SDK example, downloading the full project, and running it. img.png data ...
anliyuan/Ultralight-Digital-Human: A model that can run in real time on mobile devices; as far as I know, this should be the first open-source model that is this lightweight. ... The 20 seconds of material (actions while speaking) can be used as streaming-inference material ...
ann-yuan/QESLAT-2024: With technological progress, sign language (Sign Language Avatars) has become an important tool to promote communication between deaf communities and society; by simulating sign-language motions it provides real-time translation services and helps break language barriers ...
ascanzen/metap: This open-source project is called “Controller,” meaning it can act as the core for today’s popular virtual humans and virtual streamers (anthropomorphic digital avatars). It uses UE, C4D, DAZ, Live2D, and other 3D engine software ...
botoai/HeyGem.ai: Silicon-based AI open-source AI. Contribute to botoai/HeyGem.ai development by creating an account on GitHub.
caixukun-jinitaimei/Edubot: An education system modified from Linly-Talker, including course-summary, dialogue, and chatbot dialogue; the project can be deployed on autodl - caixukun-jinitaimei/Edubot.
chrysfay/fay-ue5-: UE5 project (Metahuman). [This is a complete open-source UE digital-human project that can be paired with the Fay framework to achieve ...
cnaiart/AIGCHUMAN: The AI system supports real-person voice and appearance cloning; you only need to upload a short real-person video to complete cloning quickly. · Based on self-developed algorithm foundations and cloud compute, it can generate content quickly, stably, at low cost. · Based on SaaS multi ...
cnaiart/Aigc-digital-human: [Sincere wrapper] The strongest “wrapper” system in the Eastern Hemisphere; frontend/backend separated; can connect to Silicon-based, Feiying, Shanjian, Yiding Open Platform, and all mainstream API interfaces; works out of the box—star it as a show of support.
data-baker/SmartGuySdkDemo: Data-Baker virtual SDK demo. Contribute to data-baker/SmartGuySdkDemo development by creating an account on GitHub.
dphoenixm/ai-virtual-human-real-time-interaction: A deployable AI virtual real-time interaction project; the virtual human has basically no latency; supported large models include GPT, Baidu Wenxin Yiyan, offline large models, Spark large model, etc.; the virtual human can run 24-hour live streaming or ...
duix-guiji/duix-sdk: ... , nice to meet you.') // drive speech with text // duix.say(' ... ', true) // drive with an audio file ...
flyarong/Fay-python-shuziren: Fay is a complete open-source project, including the Fay controller and models; it can be flexibly combined into different application scenarios: virtual streamer, on-site sales, product guide, voice assistant, remote voice assistant, interaction, interview ...
gzyxds/AIGC-SaaS_digital-human: AIGC system source code / AI SaaS system source code, aimed at business owners and individual creators building short-video IP; supports real-person voice + appearance cloning, one-click synthesis for knowledge products, courses, e-commerce selling, image promotion, ...
hailibull/AI_live: Cloning, live streaming, short-video editing SaaS; supports private self-hosted appearance cloning and voice cloning; source delivery and one year of after-sales support (renewable, worry-free); live-streaming SaaS product ...
hwhw97/fork-VideoChat: Real-time voice interaction, supporting end-to-end voice solutions (GLM-4-Voice-THG) and cascaded solutions (ASR-LLM-TTS-THG); customizable avatar and voice; no training required; supports voice cloning; first-packet latency as low as 3 seconds.
jby1993/SelfReconIntro: In recent years, with rapid advances in graphics technology, various virtual beings have entered daily life; examples include the digital astronaut Xiao Zheng, Baidu Smart Cloud AI sign-language streamer, and Tencent’s 3D sign-language digital human “Lingyu,” among others. In reality, 3D ...
jiran214/GPT-vup: Basic functions: respond to bullet chat and SC, welcome entering viewers, thank gifts; plugin (off by default). speech: listen for ctrl+t hotkey, convert speech input to text and interact with AI; action: match character actions based on viewer behavior ...
kleinlee/DH_live: Notes: the project currently mainly maintains DH_live_mini, currently the fastest 2D video solution; the project includes a web-inference example, does not rely on any GPU, and can run in real time on any mobile device.
kleinlee/MiniMates: MiniMates is a lightweight image-driven algorithm, 10–100× faster than LivePortrait, EchoMimic, and MuseTalk; it supports both voice-driven and expression-driven modes, and can be embedded in ordinary computers ...
laixiao/VirtualWifeLX: VirtualWife is a virtual project still in incubation; there is much to optimize. The author wants to build a virtual being with its own “soul”; you can get to know her like a friend; the author hopes the virtual ...
lakysir/aimh8_digital_human: Private automated queued training, short-video queued generation WeChat mini program, and one-click deployment of a web operations backend management system, based on single-person trained audio-driven lip-sync ...
libn-net/marketing_creator_pro_max_backend: High-appearance AI cloning, voice cloning, short-video generation, live streaming (to be released), AI dubbing, AI subtitles, including Windows installer, Web version, H5 version, mini program version; side-hustle essential; open-source cloning platform backend API ...
libn-net/marketing_creator_pro_max_pc: High-appearance AI cloning, voice cloning, short-video generation, live streaming (to be released), AI dubbing, AI subtitles, including Windows installer, Web version, H5 version, mini program version; side-hustle essential; open-source cloning platform Windows version and PC-Web version ...
lili3533/Fay11: Fay Controller (Is this the metaverse?) This open-source project is called “Controller,” meaning it can serve as the core for popular virtual humans and virtual streamers (anthropomorphic digital avatars). It uses ...
lipku/LiveTalking: Supports multiple models: ernerf, musetalk, wav2lip, Ultralight-Digital-Human · supports voice cloning · supports interruptible speech · supports WebRTC and virtual camera output · supports action choreography: when not speaking ...
lloves/AICameraLive: A virtual live-streaming project based on AI technology + cloud-phone technology, monetization across multiple scenarios. Contribute to lloves/AICameraLive development by creating an account on GitHub.
lyz1810/live2dSpeek: Make it speak using a Live2D model + edge-tts (text-to-speech). Contribute to lyz1810/live2dSpeek development by creating an account on GitHub.
lyzhiwang/szr-bk: Backend. Contribute to lyzhiwang/szr-bk development by creating an account on GitHub.
michael7736/digital-avatar-project: An AI-driven virtual live-streaming system supporting 2D/3D, TTS, ASR, lip-sync, streaming, interaction, and other modular development. - michael7736/digital-avatar-project.
modstart-lib/aigcpanel: AIGCPanel is a simple, easy-to-use one-stop AI system supporting video synthesis, voice synthesis, and voice cloning; it simplifies local model management and one-click import and use of AI models. aigcpanel.com ...
muttofan/DigiHM: This is a complete project including a Python core and UE model; it can be used as a digital assistant and for automated Douyin live streaming, and can also look good as an entry point for your app. - muttofan/DigiHM.
nvtb48/kreadoai: After cloning, it can even switch to other languages while still preserving your vocal timbre characteristics; for enterprises doing multilingual marketing, this is a major benefit. Not only that—there are also AI models.
oneCodeSuperman/LstmSync: LstmSync. No network required; a generalized model usable locally with 4GB VRAM! The effect seems workable; could you provide ...
shibing624/AIAvatar: AIAvatar real-time interactive streaming, achieving audio-video synchronized dialogue; basically reaches commercial quality. wav2lip. Features. Supports wav2lip lip-sync model; supports ...
shuxiaokai/Fay-AImodel: This is a complete project including a Python core and UE model; it can be used as a digital assistant and for automated Douyin live streaming, and can also look good as an entry point for your app.
sse-digital-man/TTS-Core: This project is the TTS part, providing a calling interface for the LLM part, and supplying generated audio information to the character model to generate corresponding lip-sync and actions. Architecture intro. This project provides a Web UI via ...
sugarbeby/Super-Digital-Human: AI 24-hour unlimited live streaming. Contribute to sugarbeby/Super-Digital-Human development by creating an account on GitHub.
taoofagi/easegen-admin: taoofagi/easegen-admin ; yudao-module-infra · yudao-module-infra · feat: integrate Mofa Xingyun 3D API, supporting 2D/3D dual-platform video synthesis. 3 months ago ; yudao-module-iot · yudao- ...
wan-h/awesome-digital-human-live2d: Create warmth and inject a “soul.” Community official website public beta released: https://www.light4ai.com · Bilibili video—community site introduction; the official site adds extra support on top of the open-source version (see details in the operations ...
wangCanHui/MotionVerseWeb: Motionverse open platform provides PaaS & SaaS solutions, supporting text, voice, and motion as multiple ways to drive via AI algorithms; it offers customers standard PaaS interfaces and SaaS operations tools.
way311/FayHuman: Fay is a complete open-source project, including the Fay controller and models; it can be flexibly combined into different application scenarios: virtual streamer, on-site sales, product guide, voice assistant, remote voice assistant, interaction, interview ...
willsusan/aihumanguide: Definition: an AI is a virtual character combining artificial intelligence and 3D modeling technology, able to have natural-language dialogue and interact with people. Application scenarios: customer service, education, entertainment ...
wukaikailive/digital_people: This project aims to implement a fully offline live-streaming application supporting real-time voice interaction, custom characters, a knowledge base, lip-sync, and live-stream planning. - wukaikailive/digital_people.
xaio6/Digital_Human_API: The lecture-recording system was designed and developed by the AI Horizons team as an education technology tool, aiming to provide the education sector with personalized, dedicated PPT lecture video production and presentation through advanced technology.
xaio6/Digital_Human_UI: Lecture recording system—an all-new micro-lesson video generation solution—UI. Contribute to xaio6/Digital_Human_UI development by creating an account on GitHub.
xhadmincn/GenHuman: GenHuman - a project that can be deployed for commercial monetization / GenHuman is an API-based digital human product that includes web, app, mini program, backend management, ...
xkufsxqu/deepbrain: Its most impressive part is those AIs (also called AI Avatars): the platform has built-in 2000+ licensed real-person models trained into virtual avatars, covering many skin tones, ages, genders, and professions. These ...
xp-pioneer/Fay-Assistant: Fay is a complete open-source project, including the Fay controller and models; it can be flexibly combined into different application scenarios: virtual streamer, on-site sales, product guide, voice assistant, remote voice assistant, interaction, interview ...
xszyou/Fay: The Fay framework adapts upward to various model technologies and connects downward to many LLMs; it also makes it easy to swap TTS and ASR models, providing comprehensive application interfaces for microcontrollers, apps, and websites. Changelog ...
xszyou/fay-android: The app stays resident in the phone background, so you can keep communicating with Fay anytime, anywhere. Contribute to xszyou/fay-android development by creating an account on GitHub.
xszyou/fay-ue5: UE5 project (Metahuman). [This is a complete open-source UE digital-human project that can be paired with the Fay framework to achieve ...
xtdexw/city-exhibition-guide: City showroom intelligent explanation system—an AI interactive showroom guide system based on ... Contribute to xtdexw/city-exhibition-guide development by creating an account on GitHub.
yakami129/VirtualWife: VirtualWife is a virtual project still in incubation; there is much to optimize. The author wants to build a virtual being with its own “soul”; you can get to know her like a friend; the author hopes the virtual ...
yangkang2021/I_am_a_person: 0. Data preprocessing · 1. Avatar generation and customization · 2. Input—speech recognition · 3. Brain—large language model · 4. Speaking/singing—speech synthesis · 5. Driving · 6. Deployment · 7. Other.
yuanyuekejiJN/digital-human-realtime: Windows real-time dialogue, supports local deployment, suitable for customer service, product guidance, government services, and other scenarios; works out of the box and developer-friendly. - yuanyuekejiJN/digital-human-realtime.
yuxiaolinglan/sadtalker-kaggle: SadTalker generation runnable on Kaggle; the Gradio page is modified, adding options for image restoration model selection, head movement, and other functions. - yuxiaolinglan/sadtalker-kaggle.
zhangbo2008/xuni8: 11.py # integrates codeformer for super-resolution. app8.py integrates the frontend interactive UI. Put your self-trained models in the tmp2 folder later. Summary: the dataset required for the whole project is very easy to obtain; you only need video ...
zhaoyq6/Fay-Slales: Fay is a complete open-source project, including the Fay controller and models; it can be flexibly combined into different application scenarios: virtual streamer, on-site sales, product guide, voice assistant, remote voice assistant, interaction, interview ...
Gitee (search.gitee.com) is a Git-based code hosting and collaboration platform that, in the digital human ecosystem, functions mainly as a distribution and iteration hub for Chinese-language open-source codebases and prototypes spanning “talking head” video generation, lip-sync and dubbing pipelines, speech recognition and speech synthesis integrations, LLM-driven dialogue front ends, and end-to-end application stacks such as marketing-oriented “AI digital employee” tooling and WeChat mini-program back ends; the projects surfaced under the digital human label show how developers use Gitee to publish runnable demos, forks, and localized variants (often emphasizing on-prem or LAN deployment, multi-user use, and practical workflow features like subtitles, background replacement, and batch mixing), making it a visible aggregation point where digital human implementations and product-leaning templates are shared, maintained, and adapted for China-focused platforms and operational requirements.
[51x Feb 2026]
AI-Change-Background: Digital-human background changer for images/video (replace/compose backgrounds).
ai-character-hub: Platform repo for simplifying AI digital-human creation (training... speech synthesis... interaction design).
ai-digital-human-service: Front-end and back-end code for an AI digital-human service.
AI-Vtuber: AI Vtuber Live2D virtual anchor system with LLM/TTS/stream-platform integrations.
AIGC Digital Human SaaS Source Code: SaaS multi-tenant digital human system source code (voice/image cloning + video synthesis focus).
aigcpanel: One-stop AI digital-human system (video/voice synthesis + voice cloning; local model management).
aimh8_digital_human: One-click private deployment for digital-human training + queued video generation + mini-program/backend.
ant-ai: “Ant AI” comprehensive AI application system (multimodal utilities including digital-human-related modules).
any4any: Enterprise multimodal AI platform (LLM chat + speech + knowledge base + digital-human modules).
artaigc/human: AIGC digital-human SaaS/source-code repository (business short-video IP style digital human tooling).
awesome-digital-human-live2d: Curated Live2D digital-human resources (often includes front-end interaction modes).
DH_live: Live digital-human project variant (live pipeline / service-style repository).
digital_-human: Digital-human themed profile/repo namespace (general digital human content).
digital_human: Digital-human AI project repository (general “digital human” build).
digital_human_video_player: Digital-human video player with HTTP API; typically connects to lip-sync/talking-head toolchains.
digital-human-with-llms: Digital-human project integrating LLMs (LLM-driven avatar/dialogue workflow).
digital-person: Digital-person project (general “digital human/person” implementation repository).
Double_Talker: Talking-head / two-party talker style digital-human project (dialogue-driven avatar output).
duix-js: Real-time rendering engine for digital humans (front-end rendering/runtime utilities).
duix.ai: Digital-human related project namespace (assistant/agent + avatar-facing tooling).
facechain: FaceChain model/tool for generating personal digital avatars.
Fay: Open-source digital-human framework connecting avatar layers to LLMs; modular ASR/TTS integration patterns.
fay-ue5: Fay + Unreal Engine 5 oriented adaptation (digital-human framework integration).
gomaxaipro: GoMaxAI Pro style AIGC creation system (chat + image/video/music + tooling bundle).
HenGem: HenGem digital-human batch processing system (multi-language... ops controls... model libraries).
HeyGem: Local/LAN digital-human video generator; multi-language processing with subtitles/background/montage tooling.
heygem: HeyGem repository under zhangjinyou (digital-human video generation tooling).
HeyGem.ai: HeyGem.ai mirror/sync repository (digital-human video generation toolkit).
HUAWEI MetaStudio: Huawei digital-human studio project/landing page (MetaStudio).
human: AIGC digital-human SaaS/source-code variant under xxhlq.
I_am_a_person: Real-time interactive GPT digital-human notes/project (image-based talking persona with LLM interaction).
imaiwork: IMAI.WORK (AI digital employee / automation + marketing workflow ecosystem).
KinEcho: Intelligent companion system for elderly care using a digital human (care... reminders... health management).
LHM: LHM model repository (single-view human prior + transformer output; gaussian/3D human representation focus).
Linly-Dubbing: Linly-related dubbing/speech tool repository (voice/dubbing workflow).
Linly-Talker: Digital-human dialogue system integrating LLM... ASR... TTS... and (often) voice cloning via a web UI.
Live Virtual Digital Person: Backend/service for live streaming digital human operation.
live2d: Live2D-related repository (assets/tools/implementation scaffolding).
livetalking: Real-time “live talking” digital-human/talking-head implementation (interactive streaming orientation).
LiveTalking: LiveTalking repository (real-time interactive digital-human/talking-head implementation).
LiveTalking-v15: NeRF-based real-time streaming digital human variant; often includes model/ASR add-ons.
Metahuman-Stream: Real-time streaming digital human workflow (commonly NeRF-based)... typically used for live talking-head output.
Mingyue Digital Human: Free digital-human system (upload voice... use public models... time-unlimited; TTS integration).
Mnn3dAvatar: Open-source 3D avatar/digital-human framework (face capture/mapping to 3D characters).
OpenAvatarChat: Modular interactive digital-human dialogue implementation runnable on a single PC.
ruoyi-ai: Full-stack AI platform with enterprise assistant features (often includes RAG... workflow orchestration... and digital-human hooks).
TheHeretic-VFXCharacter: Unity “Heretic” virtual/digital human VFX character repository.
Unity Digital Human: Unity-based digital human project (URP pipeline; intended for multi-platform builds).
web-digital-human: Web-oriented digital-human implementation (browser/service delivery orientation).
WGAI: WebAI training/recognition platform (digital human + OCR + speech; offline/independent deployment emphasis).
yc-digital-human: YC digital-human project repository (digital human implementation/packaging).