arkennemasis fal · MiniMax Music v2.6 · $0.15/run

Same fifteen cents as v2.5, and at a glance the two nodes look identical - same boxes, same audio settings, same price. The difference is the section vocabulary and one requirement. If you're building an arrangement rather than a vibe, this is the one to pick.

What it is

The fal wrapper for minimax-music/v2.6, which fal describes as creating complete tracks - singing, backing music, detailed arrangements - from lyrics plus a style description. One text pair in, one audio file out. It's the top of the MiniMax music family in this pack, sitting five times the price of the cheap Music v2 node, and priced identically to 2.5 right beside it.

Where the money goes is visible in the input list: a 2000-character description budget (against 300 on v2), 3500 characters of lyrics, a real set of structural tags, and an instrumental switch. It's a finishing tool, not a sketching one.

How it works

Pack-standard fal plumbing: FAL_KEY from the .env next to your launcher, input checks, a cost estimate compared against max_cost_usd, one billed request, polling while the node shows status, then download into output/fal/fal-ai_minimax-music_v2.6/. Nothing before the submit costs anything. And as with all the MiniMax nodes here - a closed, API-first lab - nothing runs on your machine; the community has basically no history with these models, so your first ten minutes of listening is the most reliable review you'll get.

Inputs

• prompt - the style description: mood, genre, scenario. 10 to 2000 characters.
• lyrics - up to 3500 characters, \n between lines. The tag list here is the reason to choose 2.6 over 2.5: [Intro], [Verse], [Pre Chorus], [Chorus], [Post Chorus], [Hook], [Bridge], [Interlude], [Transition], [Build Up], [Break], [Inst], [Solo], [Outro]. [Build Up] and [Transition] are the useful additions - they're how you get a track that moves between sections instead of jumping.
• is_instrumental - default off; on gives vocal-free music. The tooltip states lyrics are required when this is off, and doesn't offer an escape hatch the way 2.5's wording does, so write them or flip this.
• lyrics_optimizer - default off; with it on and the lyrics box empty, the model writes lyrics from your prompt.
• audio_setting_sample_rate - 16000, 24000, 32000, 44100.
• audio_setting_bitrate - 32000, 64000, 128000, 256000.
• audio_setting_format - mp3, wav, pcm.

Optional: max_cost_usd (default $20; the run stops before anything is sent when the estimate is over it, 0 removes the cap), max_concurrent (default 1 - one paid fal call at a time across a run, up to 32 for parallelism), reuse_identical_run (default on; identical inputs return the file you already paid for, even after a restart), extra_json (raw JSON merged into fal's request last, for any field this node doesn't expose).

Outputs: audio - waveform and sample rate, playable on the node - audio_path, and info with the request id, the cost estimate and fal's full answer.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/Hishamahmer/comfyui-arkennemasis
pip install httpx

Or Manager → Install via Git URL. Then:

FAL_KEY=your-key-from-fal.ai/dashboard/keys

in the .env beside run_nvidia_gpu.bat. Restart; it's under arkennemasis/fal/Audio/MiniMax.

Getting your fifteen cents' worth

Tag discipline is the whole game at this price. A lyric sheet that goes [Verse] → [Chorus] → [Verse] → [Chorus] → [Outro] with the chorus repeated word-for-word gives the model an obvious hook to build to; free-form text gives it nothing to structure. If you want a build, put [Build Up] where you want it rather than describing a crescendo in the prompt box.

Two habits from the cheap tier still apply. Draft on the 3-cent node before spending fifteen - a lyric line that sings badly will sing badly here too. And remember reuse_identical_run will return your earlier file when nothing has changed; that's the node saving you money, not failing to render.