Structured Song Prompt Source (Folder / Manual)

The toolkit's node list has two "prompt source" nodes that look nearly identical, and it's worth knowing which is which before you wire your graph. MiniMax Prompt Batch Loader is the audio-only loader. This one - Structured Song Prompt Source (Folder / Manual) - is the fuller-structured variant that also emits the artwork image_prompt, plus the full provenance trail (prompt_origin, prompt_provenance_json, variant_count). It's the node you use when you want file-driven or manual batches and a cover-art branch, but no LLM in the loop.

The tooltip is refreshingly honest about its status: "This legacy/source node does not call an LLM itself; it remains available for structured file/manual workflows and backward compatibility." So if you're starting fresh and want an LLM in the pipeline, the template + parser pair is the modern path. If you're running a deterministic file-based production chain, this node is the right shape.

Inputs

source_mode flips between folder and manual (default manual). Folder mode uses prompt_directory, extensions (.txt,.prompt,.md), and recursive. Manual mode uses manual_title, manual_caption, manual_lyrics, and - the bit the audio loader lacks - manual_image_prompt, a positive-only prompt for the cover. Keep it text/logofree if you want a text-free album cover, matching the system prompt's strict [Image_Prompt] rules.

Shared batch controls: song_count, seed_mode (random_each_song vs increment_from_base), and base_seed.

Outputs

All lists: caption, lyrics, title, image_prompt, source_name, generation_seed, run_index, variant_count, source_path, prompt_origin, prompt_provenance_json. The provenance pair is the differentiator - it records where each prompt came from (manual, file, or LLM) so the sidecar can audit the whole chain.

Installing it

Pack install via ComfyUI Manager (search "MiniMax Music Production Toolkit") or:

cd ComfyUI/custom_nodes
git clone https://github.com/jplenio/ComfyUI-MiniMax-Music-Production-Toolkit.git
cd ComfyUI-MiniMax-Music-Production-Toolkit
python -m pip install -r requirements.txt

Dependencies: scipy, soundfile, imageio-ffmpeg, mutagen, Pillow. Restart ComfyUI and hard-refresh once.

Gotchas

Same filesystem rules as every loader in this pack: directories are relative to the ComfyUI server, and unlisted extensions are skipped silently. The real decision, though, is architectural: if you're also running an LLM, don't stack this node under it - the parser already emits image_prompt and provenance. Use this node for the LLM-free branch, and keep your graph from having two sources for the same field, or you'll be debugging which one actually won.