toyxyz/ComfyUI_toyxyz_test_nodes

What it is

ComfyUI_toyxyz_test_nodes is a collection of custom nodes for ComfyUI, the visual node‑based interface used to run Stable Diffusion‑style image‑generation pipelines. The author bundles a set of handy tools they frequently need – things like webcam capture, region masks, 3‑D pose editors, video‑editing helpers, and several prompt‑generation nodes that talk to a local Qwen LLM.

Who might use it

• Artists and hobbyists who already work with ComfyUI and want ready‑made nodes for tasks such as:
  • Capturing webcam frames directly into a workflow
  • Editing 3‑D pose data or region masks
  • Building “MiniMax‑H3” audiovisual prompts for video generation
  • Turning images (or batches of up to 10) into structured Danbooru‑style prompts via a local Qwen model
• Anyone who wants to avoid writing their own Python nodes for these repetitive chores.

Core features (as described in the README)

Category	Nodes / Capabilities
Capture & Export	CaptureCam (webcam capture), Save image to path, Export glb, Export video, Direct_screenCap
Mask & Region Tools	Draw/Crop area masks, ComfyCouple Region family (region painter, multi‑region, mask, extractor)
Pose & 3‑D Editing	3‑D Pose Editor, OpenPose editor, minimax h3 camera (key‑frame camera editor)
Video Utilities	Cut Video (trim by frame count), Connect Video (concatenate with optional cross‑fade)
Prompt Generation	Minimax‑H3‑prompter (creates audiovisual prompts for the H3 pipeline), image prompter (turns images into Danbooru‑style or prose prompts using a local Qwen LLM), booru tag prompter
Miscellaneous	Load random text from file, remove noise from normal maps, depth‑to‑normal conversion, region capture, AI render overlay

Example workflow snippets

• Default workflow – shows a simple pipeline that loads an image, runs a mask node, and saves the result.
• Render preview workflow – demonstrates preview rendering inside ComfyUI.
• Direct webcam capture workflow – captures frames from a webcam without launching the separate CaptureCam UI.

How to get it running

1. Clone the repo into the custom_nodes folder of your ComfyUI installation:

   git clone https://github.com/toyxyz/ComfyUI_toyxyz_test_nodes
   

2. Run the setup.bat located in CaptureCam (sets up the webcam helper).
3. Restart ComfyUI; the new nodes appear in the node palette.

What you need to know

• Several nodes rely on a local Qwen LLM (e.g., image prompter, Minimax‑H3‑prompter). The README mentions automatic download and a managed llama.cpp runtime on first use.
• The video‑related nodes preserve original FPS and audio tracks, and they handle large video buffers by spilling to temporary files under test/cache/h3_video_memory/.
• The “MiniMax‑H3” suite is aimed at generating textual prompts for downstream video‑generation models; it does not produce the final video itself.
• All nodes are designed to be drag‑and‑drop within ComfyUI, with inputs/outputs named clearly (e.g., image_N, video_N, prompt).

Bottom line

ComfyUI_toyxyz_test_nodes is a practical toolbox for ComfyUI power‑users, giving them ready‑made nodes for webcam capture, mask creation, pose editing, video trimming/concatenation, and sophisticated prompt generation using a local Qwen LLM. It speeds up common creative pipelines without requiring users to code their own custom nodes.