Music 3.0 (BudgetPixel)

Music 3.0 is MiniMax's latest music model, and in the BudgetPixel pack it's the one to reach for first when you want a full-length song. Like its 2.6 sibling it produces complete tracks from a style prompt, sings your own lyrics if you give them, and - by default - writes or polishes the lyrics for you when you don't. It costs the same 200 credits per generation as 2.6, so "the new one is more expensive" isn't a reason to stay on the old version; the only real reason to pick 2.6 is a workflow that's already tuned around it.

It's an API node like everything in the pack. Your prompt, and optionally your lyrics, go to the BudgetPixel API; the track renders server-side; and the WAV or MP3 comes back into your graph as a core AUDIO value. No local GPU, no model download, no weights to babysit. The trade you're making is the same one the whole API-wrapper category makes - per-call cost and your inputs leaving the machine - in exchange for a model that simply doesn't have a local option.

The inputs

• prompt - style/mood description, up to 2000 characters. Optional when you supply lyrics.
• lyrics - your words, 1–3500 characters. Required unless instrumental or lyrics_optimizer is on.
• lyrics_optimizer - default true. The model writes or polishes lyrics from the prompt. Set false when your exact words are non-negotiable, or the optimizer will "improve" them.
• instrumental - vocal-free track.
• format - wav (default, lossless) or mp3.

The node exposes audio and video outputs; the video socket is a shared schema artifact that stays unconnected for audio models. audio plugs straight into Save Audio.

How it fits your workflow

The real advantage of a music node inside ComfyUI rather than a Suno tab is that the output is a graph value. Generate a track, save it, and you can wire the same audio into a lip-sync or avatar node in the pack (OmniHuman, P-Video Avatar) to have a character sing or talk over it, or drop it under a video as a soundtrack. That's the pipeline the audio-generation side of this ecosystem keeps moving toward - music isn't a dead end, it's an input to the next node.

If your budget's tighter or your lyrics are already written, the pack's Mureka V9 (60 credits, exact-lyric fidelity) and Lyria 3 (100 credits, reference-image conditioning) are cheaper. Music 3.0 is the premium all-rounder: most polish, most control, most credits.

Install and issues

Standard install - ComfyUI Manager (search "BudgetPixel") or:

cd ComfyUI/custom_nodes
git clone https://github.com/budgetpixel-ai/comfyui-budgetpixel.git
pip install -r comfyui-budgetpixel/requirements.txt

Restart, find it under Add Node → BudgetPixel → Audio, and set your API key via BUDGETPIXEL_API_KEY or a budgetpixel.json file (there's no key widget, so shared workflows stay clean).

Shared pack caveats: a 2025+ ComfyUI for the AUDIO type, and interrupting a run abandons only the local wait - the server-side job keeps running and still bills its 200 credits. And remember the lyrics-optimizer trap: on by default means your pasted lyrics can get rewritten; switch it off when exact words matter.