JoyAI Video Edit

JoyAI Video Edit

Edits a video from a plain-language instruction and streams the result back chunk by chunk while it is still being made. A demo of the JoyAI-Video-Edit model for video editors and researchers.

Made by hugging-apps on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: JoyAI Video Edit is a demo of the JoyAI-Video-Edit model. You describe an edit in plain words, and the model applies it to your clip a few frames at a time. The edited video streams back while later chunks are still being made.

The editor is a 16.3B-parameter model that handles 8 frames per chunk and keeps a rolling cache of what it has already edited. A second model, MiMo-VL-7B, turns your instruction and a frame into the conditioning.

Who it suits: Researchers and developers who work on instruction-guided video editing and want to try a streaming editor without building it.

Industry | Who uses it | How they use it | Example
Research | Video generation researchers | Test instruction-guided editing | Run a short clip with a written edit
Media | Video editors | See how a text instruction changes a clip | Try an example clip from the examples folder
Software | Developers | Judge streaming output before building on it | Watch chunks arrive while the edit runs

You can skip it if you need frame-exact control, since the edit is driven by a text instruction.

Where to use it:

• A web browser, on the Space's page at Hugging Face, with nothing to install
• Windows, macOS or Linux with Python, from a copy you download, on a large GPU

Use case | Who | What they get back
Edit a clip with words | Editors | The clip edited to match your instruction
Watch streaming output | Researchers | Edited chunks arriving while generation runs
Try the sample clips | Newcomers | Showcase clips with the authors' prompts

How to set it up:

• Python 3.12 on your computer
• The Python packages listed in requirements.txt
• A large GPU, since the Space fits the model on a 48 GB slice
• The huggingface_hub package, which gives you the hf command

1. Install the huggingface_hub package.
2. Download the Space with the hf command.
3. Move into the downloaded folder.
4. Install the pinned Gradio version.
5. Install the packages from requirements.txt.
6. Start the app with Python.

pip install -U huggingface_hub
hf download hugging-apps/jdopensource-joyai-video-edit --repo-type space --local-dir jdopensource-joyai-video-edit
cd jdopensource-joyai-video-edit
pip install gradio==6.22.0
pip install -r requirements.txt
python app.py

How to use it: Open the Space's page in your browser, or start your own copy with the command below. Add a clip, write the edit you want in plain language and run it. The defaults follow the reference server's settings.

python app.py

Example: You add a short clip and write the change you want in a sentence. The demo edits the clip chunk by chunk, and you see the edited video stream back while it is still being generated.

For newcomers and experts: If you are new to AI video editing, open one of the example clips and run it with the authors' prompt first. Then try a small change of your own on a short clip.

If you know the field, the Space is a port of the reference deployment with changes for ZeroGPU. It uses PyTorch attention instead of FlashAttention, pure PyTorch for fused operations, torchao for FP8 and no torch.compile. Stages run in order rather than in parallel threads, and the face restoration gate is left out.

You add a clip and write an edit
The condition encoder reads the instruction
The editor changes 8 frames at a time
The decoder returns each chunk
The edited video streams back to you

Good to know:

• License: the Space lists apache-2
• Cost to run: none named, though it runs on ZeroGPU hardware at Hugging Face
• Data it reaches: the clips you add, plus the model weights and the sample clips in the Space
• Not for: frame-exact edits and face restoration, which the Space leaves out

On Hugging Face, 13 November 2026

• Likes: 13
• Built with: gradio
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/hugging-apps/jdopensource-joyai-video-edit
• Files: https://huggingface.co/spaces/hugging-apps/jdopensource-joyai-video-edit/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.