Vidu S

Vidu S2 extends real-time video generation beyond talking-head digital characters to high-resolution interactive avatars, live video editing, and immersive spatial video. It includes Vidu S2-Avatar for controllable character generation and Vidu S2-Editing for transforming live video streams.

Key breakthroughs:

1. 720p real-time interactive avatars
   • Vidu S2-Avatar generates 720p video at 25–42 FPS, supports complex instructions such as dancing, and accepts new reference images at any moment during streaming.
2. Stable long-horizon generation with Self-Replay Forcing
   • Self-Replay Forcing replays re-noised, self-generated trajectories during training to reduce error accumulation across streaming segments.
3. Real-time editing of live video
   • Vidu S2-Editing transforms incoming video streams using reference images, supporting style transfer, virtual try-on, character replacement, and background replacement while preserving source motion.
4. Real-time spatial video for immersive displays
   • Produces synchronized stereo views for immersive displays and VR headsets.
5. Efficient inference on low-cost GPUs
   • TurboDiffusion and TurboServe combine efficient attention, low-bit GEMM, kernel optimizations, and multi-GPU pipelining for real-time inference.

Vidu S1 is a real-time interactive video generation model for voice-controlled digital characters. Users can guide generated video content at any moment through spoken instructions, enabling live interaction.

Key breakthroughs:

1. Real-time speech control over video content
   • Users can directly instruct digital characters to perform actions.
2. Infinite-length real-time interactive generation
   • Vidu S1 generates 540p video at up to 42 FPS and can run on consumer GPUs.
3. Custom character images and voice tones
   • Vidu S1 supports real people, anime-style characters, pets, and other personalized avatars.

• Try Vidu S2: https://www.vidu.com/vidu-stream
• Vidu S2 Paper: https://arxiv.org/abs/2609.11638

• Vidu S1 Paper: https://arxiv.org/abs/2607.03118

• User Guide: shengshu.feishu.cn/wiki/X7ZLwyLUzi461LkAlNpcTxs1nsy
• API Documentation: platform.vidu.com/vidu-stream/doc
• Vidu S2-Avatar Quickstart: platform.vidu.com/vidu-stream/doc/s2-avatar/realtime/quick-start
• Vidu S2-Editing Quickstart: platform.vidu.com/vidu-stream/doc/s2-editing/quick-start

• User Guide: shengshu.feishu.cn/wiki/Fz8ywkzEwil84LkbCOKcxotMnvb
• API Documentation: platform.vidu.cn/vidu-stream/doc
• Vidu S2-Avatar Quickstart: platform.vidu.cn/vidu-stream/doc/s2-avatar/realtime/quick-start
• Vidu S2-Editing Quickstart: platform.vidu.cn/vidu-stream/doc/s2-editing/quick-start

For agent-assisted API integration, use the vidu-s-api Skill. In Claude Code, Codex, OpenClaw, or any agent that supports Skills, say directly:

Install this skill: https://github.com/shengshu-ai/vidu-s-api/tree/main/skills/vidu-s-api

The agent will clone and install it into the proper Skills directory. Restart the agent if required, then ask it to load vidu-s-api for Vidu S2 API integration.

• [2026-09]: Vidu S2 is now available to try at vidu.com/vidu-stream
• [2026-07]: Vidu S1 is now available.

If you find Vidu S useful for your research, please cite:

@article{zhang2026vidus2,
  title={Vidu S2: Real-Time Interactive, Editable, and Spatial Video Generation},
  author={Zhang, Jintao and Jiang, Kai and Chen, Jintao and Wang, Xu and Liu, Deyuan and Li, Jungang and Chen, Dechuang and Lin, Ming and Zhou, Jingjiang and Jin, Haopeng and others},
  journal={arXiv preprint arXiv:2609.11638},
  year={2026}
}

@article{zhang2026vidus1,
  title={Vidu S1: A Real-Time Interactive Video Generation Model},
  author={Zhang, Jintao and Jiang, Kai and Chen, Jintao and Wang, Xu and Luo, Yang and Wang, Yuji and Chen, Dechuang and Li, Jungang and Ye, Chengyang and Chen, Marco and others},
  journal={arXiv preprint arXiv:2607.03118},
  year={2026}
}