Open-Source Frontier Voice AI

Neural building blocks for speaker diarization: speech activity detection, speaker change detection, overlapped speech detection, speaker embedding

Use PEFT or Full-parameter to CPT/SFT/DPO/GRPO 600+ LLMs (Qwen3.5, DeepSeek-R1, GLM-5, InternLM3, Llama4, ...) and 300+ MLLMs (Qwen3-VL, Qwen3-Omni, InternVL3.5, Ovis2.5, GLM4.5v, Llava, Phi4, ...)…

A Fundamental End-to-End Speech Recognition Toolkit and Open Source SOTA Pretrained Models, Supporting Speech Recognition, Voice Activity Detection, Text Post-processing etc.

Agent framework and applications built upon Qwen>=3.0, featuring Function Calling, MCP, Code Interpreter, RAG, Chrome extension, etc.

Qwen3.5 is the large language model series developed by Qwen team, Alibaba Cloud.

[CVPR2024] DisCo: Referring Human Dance Generation in Real World

Enjoy the magic of Diffusion models!

🏂 Training-Free Human Mesh Recovery from Videos, based on SAM-3, Diffusion-VAS, and SAM-3D-Body.

Qwen3-VL is the multimodal large language model series developed by Qwen team, Alibaba Cloud.

Wan: Open and Advanced Large-Scale Video Generative Models

TurboDiffusion: 100–200× Acceleration for Video Diffusion Models

SoulX-FlashTalk is the first 14B model to achieve a sub-second start-up latency (0.87s) while sustaining a real-time throughput of 32 FPS

[AAAI 2026] EchoMimicV3: 1.3B Parameters are All You Need for Unified Multi-Modal and Multi-Task Human Animation

rCM: SOTA Diffusion Distillation & Few-Step Video Generation based on sCM/MeanFlow

Multilingual Voice Understanding Model

Implementation of "Live Avatar: Streaming Real-time Audio-Driven Avatar Generation with Infinite Length"

Krea Realtime 14B. An open-source realtime AI video model.

​​Unlimited-length talking video generation​​ that supports image-to-video and video-to-video generation

开源的LstmSync数字人泛化模型，只做最好的泛化模型！

Wan: Open and Advanced Large-Scale Video Generative Models

【Accepted by TPAMI】Human Motion Video Generation: A Survey (https://ieeexplore.ieee.org/document/11106267)

[NeurIPS 2025 D&B🔥] OpenS2V-Nexus: A Detailed Benchmark and Million-Scale Dataset for Subject-to-Video Generation

HuMo: Human-Centric Video Generation via Collaborative Multi-Modal Conditioning

We present StableAvatar, the first end-to-end video diffusion transformer, which synthesizes infinite-length high-quality audio-driven avatar videos without any post-processing, conditioned on a re…

Taming Stable Diffusion for Lip Sync!

[ICLR 2025 Oral] TANGO: Co-Speech Gesture Video Reenactment with Hierarchical Audio-Motion Embedding and Diffusion Interpolation

[CVPR 2025] Hallo3: Highly Dynamic and Realistic Portrait Image Animation with Video Diffusion Transformer

Original reference implementation of "3D Gaussian Splatting for Real-Time Radiance Field Rendering"