Prism
by Tencent-Hunyuan
Prism is a research framework from Tencent Hunyuan and Fudan University for training AI models that generate video and matching audio together at up to 2K resolution. It uses a dynamic sparse attention method that makes training about 2.5 times faster than full attention, and it includes training code, inference code and a preview model.
- Stars
- 1.5k +36 a day, est.
- Licence
- MIT
- Runs on
- A gpu machine, with a GPU
- Setup
- Technical
- Language
- Python
- Age
- 41 days
What you can do with it
- Generate video with synchronised audio at 720p, 1080p or 2K using the preview model checkpoint
- Fine-tune or train a joint video-audio model with the supplied training and full finetuning code
- Pre-process your own footage into latents and decode them again for training
Make money with it
An individual with suitable GPU access could offer custom AI-generated video clips with audio to small brands, though the licence terms need checking first.
An idea, not financial advice. Check the licence before charging for anything built on it.
Ways a business could use it
- Produce short promotional video clips with built-in sound without paying per-clip fees to a hosted video service
- Fine-tune a model on your own brand footage to generate consistent product or social media videos
Can you use it commercially?
Free to use, change and sell (MIT-style terms in the licence file). Some folders may carry other terms.
Not legal advice. Read the project's licence file before building a product on it.
Getting it running
Install Python 3.10 or later, CUDA 12.4 or later, PyTorch, FlashAttention and ffmpeg with pip and apt, then download the model weights from Hugging Face.
Follow the official instructions in the project README.
Things to check
- Licence is not clearly stated, so commercial use is uncertain
- Needs a powerful NVIDIA GPU with CUDA 12.4 or later, and 2K training likely needs a multi-node cluster
- Only a preview model is released; the Prism-pro version is not yet available
- Research code with a demanding setup, not a ready-made app
Who's talking about it
- findarepo: 30 Best AI Video Generator GitHub Repos (October 2026) — Open-Source Tools Featured in a roundup
Checked 02 Nov 2026. Summary drafted by AI from the project's README.
