# HunyuanVideo 1.5 — card for AI agents

8.3B-parameter text-to-video and image-to-video model from Tencent that runs on consumer GPUs.

Category: AI video generators and platforms · maker Tencent Hunyuan · checked 2026-10-25

## Price
- Price: Free
- Free tier: Open weights, free to download; not licensed in the EU, UK or South Korea.

## License
Tencent Hunyuan Community License — https://github.com/Tencent-Hunyuan/HunyuanVideo-1.5/blob/master/LICENSE

- commercial: **conditions** (Commercial use with conditions)
  > THIS LICENSE AGREEMENT DOES NOT APPLY IN THE EUROPEAN UNION, UNITED KINGDOM AND SOUTH KOREA AND IS EXPRESSLY LIMITED TO THE TERRITORY, AS DEFINED BELOW.
  (https://github.com/Tencent-Hunyuan/HunyuanVideo-1.5/blob/master/LICENSE)
- attribution: **no** (No credit needed)
  > You are encouraged to: (i) publish at least one technology introduction blogpost or one public statement expressing Your experience of using the Tencent Hunyuan Works; and (ii) mark the products or services developed by using the Tencent Hunyuan Works
  (https://github.com/Tencent-Hunyuan/HunyuanVideo-1.5/blob/master/LICENSE)
- in app: **conditions** (Ship inside your app with conditions)
  > the monthly active users of all products or services made available by or for Licensee is greater than 100 million monthly active users in the preceding calendar month, You must request a license from Tencent
  (https://github.com/Tencent-Hunyuan/HunyuanVideo-1.5/blob/master/LICENSE)
- youtube: **conditions** (Monetized YouTube with conditions)
  > You must not use, reproduce, modify, distribute, or display the Tencent Hunyuan Works, Output or results of the Tencent Hunyuan Works outside the Territory.
  (https://github.com/Tencent-Hunyuan/HunyuanVideo-1.5/blob/master/LICENSE)

## Rules to follow
- Not licensed in the EU, UK or South Korea; the weights and outputs may not be used there.
- Products with over 100M monthly active users need a separate license from Tencent.
- Ship a Notice file with any redistribution; outputs may not be used to improve other AI models.
- Follow the Acceptable Use Policy attached to the license.

## How to connect
- Download only
- Repository: https://github.com/Tencent-Hunyuan/HunyuanVideo-1.5
- Website: https://huggingface.co/tencent/HunyuanVideo-1.5

## Notes
Hugging Face API, Oct 25, 2026: 1352 downloads, 1054 likes; GitHub Tencent-Hunyuan/HunyuanVideo-1.5: 4578 stars.

## Numbers (AgentGid data)
- Gid Score: 25
- GitHub stars: 4,609

## Lines to paste into your prompt or AGENTS.md

```
- HunyuanVideo 1.5 (AI video generators and platforms). 8.3B-parameter text-to-video and image-to-video model from Tencent that runs on consumer GPUs. Price: free; commercial use with conditions; no credit needed; ship inside your app with conditions.
  Connect: download from https://huggingface.co/tencent/HunyuanVideo-1.5 (no API).
  Rule: Not licensed in the EU, UK or South Korea; the weights and outputs may not be used there.
  Rule: Products with over 100M monthly active users need a separate license from Tencent.
  Rule: Ship a Notice file with any redistribution; outputs may not be used to improve other AI models.
  Rule: Follow the Acceptable Use Policy attached to the license.
  License: https://github.com/Tencent-Hunyuan/HunyuanVideo-1.5/blob/master/LICENSE (checked 2026-10-25; details: https://agentgid.com/assets/hunyuanvideo/)
```

## Links
- AgentGid page: https://agentgid.com/assets/hunyuanvideo/
- Is it free for commercial use? https://agentgid.com/assets/hunyuanvideo/commercial-use/
- Source: https://huggingface.co/tencent/HunyuanVideo-1.5
- Source: https://github.com/Tencent-Hunyuan/HunyuanVideo-1.5
- Source: https://github.com/Tencent-Hunyuan/HunyuanVideo-1.5/blob/master/LICENSE


## HunyuanVideo 1.5: open video model for home graphics cards

HunyuanVideo 1.5 is an open AI model from Tencent that makes short videos from a written description or brings a still picture to life. It is small enough to run on one home graphics card with at least 14 GB of memory. Its license does not allow use in the European Union, the United Kingdom or South Korea.

### Where to use it
- A short clip from a description: Write a detailed sentence about the scene and get a 480p or 720p clip of about five seconds. Longer, richer descriptions give better videos.
- Bring a photo to life: Start from your own picture, such as a product shot or a drawing, and describe the movement you want. A faster model does this in about a minute on an RTX 4090.
- Make videos from Python code: Call the model from a Python script, for example to make a batch of clips for a catalog or a content plan.
- Sharper 1080p results: Turn the finished clip into a sharper 1080p version with the built-in upscaling step, for example for a website header.
- Teach it your own style: Train the model further on your own clips, for example a brand's look, with the released training code and small add-on files.
- Short ads with readable text: Make clips where a sign or a sheet of paper shows a short word or slogan, which many video models draw badly. (idea, not from the official pages)

### How to start
- Get the code on a Linux machine with an NVIDIA card and install the packages.

```
git clone https://github.com/Tencent-Hunyuan/HunyuanVideo-1.5.git
cd HunyuanVideo-1.5
pip install -r requirements.txt
```

- Download the model files, then the text readers described in the download guide.

```
hf download tencent/HunyuanVideo-1.5 --local-dir ./ckpts
```

- Run generate.py with your prompt, a resolution and the model folder; leave the image path as none for text only.

```
python generate.py --prompt "your prompt" --image_path none --resolution 480p --rewrite false --model_path ./ckpts
```


### License
Under the Tencent Hunyuan Community License you may use it and its videos, also commercially, but not in the EU, the UK or South Korea. Products with over 100 million monthly users need Tencent's permission, and outputs may not train other AI models.

### Good to know
- The full download is very large. Each model variant is about 33 GB, and the whole set is over 350 GB.
- The setup guide also downloads a vision encoder from the FLUX.1-Redux-dev page, which needs an approved access request and a personal access token.
- That FLUX.1-Redux-dev page is marked with a non-commercial license, so check it before paid work.
- Prompt rewriting is on by default and sends your prompt to a separate language model server, if you set one up.
- Videos posted publicly must be clearly marked as machine-made.

Checked 2026-11-12 against: https://huggingface.co/tencent/HunyuanVideo-1.5, https://huggingface.co/tencent/HunyuanVideo-1.5/raw/main/README.md, https://huggingface.co/api/models/tencent/HunyuanVideo-1.5/tree/main?recursive=true, https://github.com/Tencent-Hunyuan/HunyuanVideo-1.5, https://github.com/Tencent-Hunyuan/HunyuanVideo-1.5/blob/master/README.md, https://github.com/Tencent-Hunyuan/HunyuanVideo-1.5/blob/master/LICENSE, https://github.com/Tencent-Hunyuan/HunyuanVideo-1.5/blob/master/requirements.txt, https://github.com/Tencent-Hunyuan/HunyuanVideo-1.5/blob/master/checkpoints-download.md, https://huggingface.co/api/models/black-forest-labs/FLUX.1-Redux-dev

### Mistakes to avoid
- Don't: hf download tencent/HunyuanVideo-1.5 for a first try on a small disk → Do: download only the folders you need, for example transformer/480p_t2v, vae and the text encoders (Every model variant is about 33 GB, so the full set fills hundreds of GB.)
- Don't: using the cfg-distilled model with a few steps to save time → Do: run cfg-distilled models with 50 steps; only the step-distilled 480p model uses 8 or 12 (The makers say the cfg-distilled models give wrong results otherwise.)
- Don't: stopping at an out-of-memory error on a card with more than 14 GB → Do: export PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True,max_split_size_mb:128 and run again (The README gives this setting as the fix.)
- Don't: --sparse_attn true on an RTX card → Do: leave sparse attention off unless you have an H-series card (Sparse attention needs H-series cards.)
- Don't: setting it up for a person or company in the EU, UK or South Korea → Do: stop and pick a model without territory limits, such as Wan 2.2 (The license does not apply there, outputs included.)

### Questions
- Can I use it from the EU, the UK or South Korea? No. The license does not apply there, and the model and its videos may not be used or shown there.
- Who owns the videos I make? Tencent claims no rights in them; you and your users are responsible for them.
- Which graphics card do I need? An NVIDIA card with at least 14 GB of memory when offloading is on; more memory runs faster.
- Does it work in ComfyUI? Yes, ComfyUI supports it and the project has a ComfyUI guide.

---
AgentGid's catalog is curated: every entry was checked against the maker's official pages (registry, repository, license and pricing pages) on the date shown; a license value shows the quote it stands on wherever AgentGid has one. Usage numbers come from AgentGid's own scan of public repositories set up for AI coding agents. Claims by makers are reported, not tested by AgentGid.
