AI dataset preparation toolkit for diffusion model LoRA training.
1.4K
AI dataset preparation toolkit for diffusion model LoRA training.
aitools provides 3 command-line tools and Python APIs for preparing image and video datasets. Also included is a tool called chararep which uses the output of the other tools to swap faces in videos.
| Tool | Command | Description |
|---|---|---|
| Portrait Prep | portrait-prep | End-to-end portrait image preparation (convert → crop → caption → augment) |
| Video Crop | vicrop | Extract face-cropped PNG frames from video files |
| Video Description | videsc | Generate text descriptions for video files — fast WD14 tag-based captions (default) or rich natural-language descriptions via Qwen3-VL / Qwen3-omni / Qwen3.5 (--vl) or Gemma4 (--gemma4) or a remote vLLM server (--vllm) |
| Character Replace | chararep | Replace character faces in a video using deep face-swapping models (inswapper, SimSwap, uniface, hyperswap, blendswap). Other tools like vicrop and portrait-prep can be used to create the portrait galleries required by chararep |
This docker image matches the github repo here: https://github.com/otherroch/aitools
This simple docker command can be used to run it:
docker run -it --gpus all otherroch/aitools bash
Content type
Image
Digest
sha256:42b84fa46…
Size
12.9 GB
Last updated
4 months ago
docker pull otherroch/aitools