kjuma-omni-bundle

Sub-packages bundled into a single repo so ComfyDeploy (and any other ComfyUI deploy platform) can install everything with one Git URL.

Included sub-packages

• sprite_sheet_router/ — see its own __init__.py for NODE_CLASS_MAPPINGS
• omni_animation_sheet/ — see its own __init__.py for NODE_CLASS_MAPPINGS
• omni_game_audio_midi/ — see its own __init__.py for NODE_CLASS_MAPPINGS
• omni_pose_template_refs/ — see its own __init__.py for NODE_CLASS_MAPPINGS
• omni_sfx_master/ — see its own __init__.py for NODE_CLASS_MAPPINGS
• comfyui_pixal3d_preprocess/ (renamed from ComfyUI-Pixal3D-Preprocess — Python identifiers can't contain -) — see its own __init__.py for NODE_CLASS_MAPPINGS
• comfyui_multi_provider_prompt/ (renamed from comfyui-multi-provider-prompt — Python identifiers can't contain -) — see its own __init__.py for NODE_CLASS_MAPPINGS
• vibevoice_comfyui/ (renamed from VibeVoice-ComfyUI — Python identifiers can't contain -) — see its own __init__.py for NODE_CLASS_MAPPINGS

How ComfyDeploy uses it

When this repo's clone is dropped into ComfyUI/custom_nodes/, ComfyUI loads the root __init__.py which imports each sub-package and merges their NODE_CLASS_MAPPINGS + NODE_DISPLAY_NAME_MAPPINGS. Result: every node appears in the UI / /object_info as if it had been installed separately.

Recovering excluded assets

omni_game_audio_midi/vendor/Text2midi/captions/captions.json (66 MB, training-only dataset) is not in the bundle. The code only needs it when re-training, never during inference. If you ever need it back:

cd omni_game_audio_midi/vendor/Text2midi
curl -L https://github.com/AMAAI-Lab/Text2midi/raw/main/captions/captions.json -o captions/captions.json

License

MIT (see LICENSE). Each sub-package retains its own internal license where applicable.