Suno v6 Mini · Text to Audio

Here's the thing that surprised me about this node: Mini costs the same as full v6. 10.8 credits a run, exactly. Mini isn't the cheap tier in this pack - it's the fast tier. PopNow's own summary sells it as "two quick songs with vocals from one sentence", and "quick" is the operative word.

So don't install it as a budget option. Install it as the thing you run before the real one.

What it's for

Suno's full v6 wants a brief: genre, mood, tempo, instruments, voice. Mini wants a sentence. That's a different workflow - you're not specifying a production, you're throwing a concept at a wall to hear whether the concept has legs.

I use it like this: one line in prompt, generate_lyrics left on, two versions back for eleven cents. If one of them makes me sit up, I take the style description Mini inferred, write it properly, and run full v6. If both are boring, I've learned that in a couple of minutes instead of a couple of hundred seconds.

Which is honest about the trade: on a mixed job the mini run is a cost, not a saving. It pays for itself the first time it stops you spending a full run on a bad idea.

The inputs

The schema is identical to the full v6 node, field for field - prompt, lyrics, generate_lyrics, instrumental, title, negative_prompt, vocal_gender, style_weight, weirdness, audio_weight, variety, persona_id, persona_model, plus the three optional autogrow reference sockets. Nothing is missing. The difference is the model behind it, and how much it will do with a thin brief.

prompt is where you'll live. One sentence: genre, mood, what it's for. lyrics empty, generate_lyrics on. If you have lyrics, this drops back to being a normal song node - which is fine, it just isn't playing to its strength.

The section tags are worth knowing even in quick mode: [Verse], [Chorus], [Bridge] mark structure and are not sung. The prompt field caps at 1,000 characters, which is more style description than a one-sentence brief needs.

variety still controls how different the two versions are from each other (0 off, 1 normal, 2 high, 3 extra, 4 max), and it defaults to -1 - meaning the field is left out of the request and the model decides. On the quick path, leave it. You want the two takes to diverge; that's the audition.

style_weight and weirdness run 0 to 1, and both default to -1 for "model's choice". Reach for weirdness only when the quick result came back too polite.

The reference sockets - up to five image_urls, one reference_video_urls clip (100 MB, 241 seconds max), one reference_audio_urls clip (6 seconds to 30 minutes) - only apply when the model writes the lyrics. Mini with a handful of mood stills is a surprisingly effective way to get words out of a vibe.

Outputs: audio (a batch - Save Audio writes one file per song) and task id.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/popnow-ai/popnow-comfyui

Or Manager → search PopNow → Install → restart. Key from popnow.ai/api/console/keys in Settings > PopNow, or:

POPNOW_API_KEY=sk-popnow-... python main.py

No dependencies beyond what ComfyUI ships, no model fetch, no GPU. Restart again after adding the key - the node list is built from your account catalog at startup.

Where people get burned

It is not the cheap one. If you're picking between Mini and v6 on price, there's nothing to pick. Pick on speed versus how much control you need.

Quick songs are short and simple. One sentence in, a two-minute pop structure out. Ask Mini for a seven-minute prog track with a key change and you get a normal song that ignores you. That's the moment to move up to v6, or to Wild if the brief is strange rather than complex.

No seed. Same as the rest of the Suno nodes - no reproducibility, so save what works. The task id is the only way back to a run, via PopNow's API console.

Cancelling isn't a refund. Interrupting stops ComfyUI waiting and nothing else; the job finishes on PopNow and is charged if it succeeds.