AQ_parse_json_to_acestep

The acestep family's secret weapon: a pure parser that takes a JSON string - from any LLM node, not just this pack's - and splits it into the exact outputs the ACE-Step pipeline expects. No API key, no network, no model, no cost. If you already have a local model or another LLM node producing JSON, this is how you plug it into the song-generation flow.

Why it exists

The pack's own LLM nodes (AQ_Gemini_acstep15, AQ_OpenAI_acstep15, AQ_openai_compatible) all return the same 16-slot metadata set - tags, lyrics, seed, bpm, duration, timesignature, language, keyscale, coverImage, title, tagline, headline, and the three yt_* fields. But maybe you want the metadata from a different tool, or you'd rather call the LLM yourself. This node decouples the LLM from the music pipeline: JSON in, typed slots out.

How it works

It regex-extracts the first {...} block from the input - so markdown-wrapped JSON from sloppy LLMs still parses - then json.loads it and coerces each field to its output type with defaults: bpm→120, duration→120.0, timesignature→"4", language→"en", keyscale→C major. Literal \n escapes in lyrics get unescaped into real newlines.

Inputs

• json_text - the JSON string, multiline.

Outputs

The 16 acestep slots: response_json, tags, lyrics, seed, bpm, duration, timesignature, language, keyscale, coverImage, title, tagline, headline, yt_description, yt_tags, yt_title.

Install

Ships in AQnodes:

cd ComfyUI/custom_nodes
git clone https://github.com/2frames/ComfyUI-AQnodes
cd ComfyUI-AQnodes
pip install -r requirements.txt

or search "AQnodes" in ComfyUI Manager and restart.

Gotchas

It expects the exact schema keys the pack's LLM nodes produce - bpm, durationSeconds, keyscale, and so on. JSON from a random node won't populate the slots unless it uses those same keys; a field your LLM calls "tempo" instead of "bpm" just silently defaults. And unlike the API nodes, it raises on malformed JSON rather than returning empties - check the console if nothing comes out. Honest bottom line: if you're already using one of the pack's acestep LLM nodes, you don't need this one. It's for when you're not.