Skip to content

feat: add MuScriptor Load Model node for manual model selection - #2

Open
moonwhaler wants to merge 1 commit into
martyyz-ai:mainfrom
moonwhaler:feat/manual-model-loading
Open

feat: add MuScriptor Load Model node for manual model selection#2
moonwhaler wants to merge 1 commit into
martyyz-ai:mainfrom
moonwhaler:feat/manual-model-loading

Conversation

@moonwhaler

@moonwhaler moonwhaler commented Aug 15, 2026

Copy link
Copy Markdown

Summary

  • Adds a MuScriptor Load Model node that outputs a MUSCRIPTOR_MODEL, feeding into MuScriptor Transcribe Audio to MIDI.
  • model_source dropdown offers auto (download): small/medium/large (existing HuggingFace auto-download behavior) plus any local .safetensors found in a new ComfyUI/models/muscriptor/ folder — so users with an already-downloaded checkpoint no longer need to re-authenticate/re-download.
  • Updates README (EN/PT) and the example workflow to match.

Closes #1

Test plan

  • python -m py_compile comfy_node.py
  • Validated example workflow JSON parses
  • Manual test in a running ComfyUI instance (loader node shows local files + auto-download options, Transcribe node runs with connected model)

Replaces the model_size/device/dtype widgets on the Transcribe node with
a dedicated loader node that outputs a MUSCRIPTOR_MODEL. Lets users drop
their own .safetensors checkpoint into ComfyUI/models/muscriptor/ instead
of only auto-downloading published HuggingFace variants.
@moonwhaler moonwhaler mentioned this pull request Aug 15, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

models path

1 participant