Suno v6-wild is less predictable: how to keep a track on Sume

On Sume you cannot regenerate the same track on demand, because Music 1.0 exposes no seed and no temperature; the only repeatable thing is the audio file a finished job returned. Suno's v6 post describes v6-wild as "less predictable and more varied" than v6, which is the same lesson from the other side: a good result is a one-off, so store it.

Sources: Suno's v6 announcement (read 2026-10-10) and Sume's Music 1.0 docs.

What controls does Sume Music give you?

The documented inputs are prompt (1 to 5,000 characters) and an optional public HTTPS image_url, where null clears it. There is no seed, no temperature and no duration field. The docs recommend writing a seven-axis brief (genre and mood, instrumentation, tempo and energy, structure, production feel, use case, constraints) and adding "Instrumental, no vocals" when you want no singing.

• Same prompt twice gives two different tracks, and both are billed at $0.125.
• result.lyrics is model-reported metadata, not a transcript you can rely on.
• Artifacts are audio/mpeg; download and store your own copy.

How do you make a result repeatable?

Treat the first accepted track as the master. Save the artifact file to your own storage, and record the request body plus, for the router, job.request.routed_model, which names the engine that actually ran. job.model echoes what you asked for, so with sume/music-auto the two can differ and the second is the one to log.

When does variety help?

If you want options, variety is the point: generate several tracks, listen, and keep one. Budget it: five takes cost $0.625, ten cost $1.25. Because a policy rejection returns an error rather than a track, change only the flagged part of the brief and keep the musical direction, as the docs advise.

What about Suno's own tiers?

Suno says v6-wild is for Pro and Premier users. Whether its output is easier to repeat is not stated on the page read here, and Sume does not list Suno, so use Suno's tools directly if you need a Suno result and bring the finished file into Sume for captions, timeline mixing or a lip-sync clip.

A small script for a keep-everything habit

Whatever you generate, copy the audio the day you make it. Result URLs are for fetching; your own storage is the archive. A short loop that downloads each finished track and writes the request beside it takes a few lines and saves the argument later about which take was approved.

Name files with the date and a short slug of the brief, and keep a CSV row with takes generated and takes kept. After a month you will know your real keep rate, which is the number that decides whether variety is worth it.

• Download the file; do not rely on the URL staying put.
• Keep a row per brief, not per take.
• Record who approved which take.