45 lines
1.4 KiB
Markdown
45 lines
1.4 KiB
Markdown
# Contributing
|
|
|
|
Thanks for your interest in improving ComfyUI-VideoToText.
|
|
|
|
## Dev setup
|
|
|
|
```bash
|
|
# In your ComfyUI install:
|
|
cd ComfyUI/custom_nodes
|
|
git clone https://github.com/aadebuger/ComfyUI-VideoToText.git
|
|
cd ComfyUI-VideoToText
|
|
|
|
# Install deps into ComfyUI's venv
|
|
source ../../.venv/bin/activate # adjust to your ComfyUI venv path
|
|
uv pip install -r requirements.txt
|
|
```
|
|
|
|
Then restart ComfyUI and look for the `🎙️ VideoToText` category in the node search.
|
|
|
|
## Reporting bugs
|
|
|
|
Please include:
|
|
|
|
- ComfyUI version (`git rev-parse HEAD` inside the ComfyUI repo)
|
|
- PyTorch + CUDA versions (`python -c "import torch; print(torch.__version__, torch.version.cuda)"`)
|
|
- `faster-whisper` version (`pip show faster-whisper`)
|
|
- `ffmpeg -version` output
|
|
- The minimal workflow JSON that reproduces the issue (use the CLI's `--dump-workflow` flag)
|
|
- Full stderr / ComfyUI console traceback
|
|
|
|
## Pull requests
|
|
|
|
- Keep nodes single-purpose. New node ideas: SRT/VTT exporter, transcript diff, translation, summarization.
|
|
- Don't reach into ComfyUI internals beyond `folder_paths`.
|
|
- Lazy-import heavy deps (`faster_whisper`, `torch`) inside method bodies, not at module top.
|
|
- Preserve the API contract for existing nodes — add new optional fields, don't rename required ones.
|
|
- Run the example workflow before opening the PR.
|
|
|
|
## Style
|
|
|
|
- Type hints encouraged but not enforced.
|
|
- f-strings, not `.format()`.
|
|
- 4-space indent, no tabs.
|
|
- One node per file under `nodes/`.
|