Lab: Does v6-wild really drift? A genre-adherence test across Suno v6 models
We plan to run one fixed brief through v6, v6-wild and v6-mini with Variety off to measure genre and tempo adherence. The findings below are mock.
Hypothesis
Suno describes v6-wild as pushing "further from the prompt." With Variety set to Off, we expect v6 to hit the requested genre and BPM most often, v6-wild least often, and v6-mini in between with thinner arrangements.
Setup
- Models: v6, v6-wild, v6-mini. 10 generations each (30 total).
- Controls: Variety Off, Style Influence 50% (default), Weirdness default, instrumental off, same lyrics.
- Styles prompt:
1970s Ethiopian jazz, 96 BPM, minor pentatonic horn riffs, vibraphone,
fuzzy organ, tight funk drums, male baritone vocal in English, warm tape saturation
Exclude: EDM, trap hi-hats, autotune- Lyrics: 2 verses and a chorus with
[Verse]/[Chorus]tags. - Scoring: BPM measured with a tempo detector (±3 BPM counts as a hit), blind genre-match votes from 3 listeners (1–5 scale), and whether any excluded element appears.
Mock findings (illustrative only, not real results)
| Metric | v6 | v6-wild | v6-mini |
|---|---|---|---|
| BPM within ±3 | 8/10 | 5/10 | 7/10 |
| Mean genre-match score | 4.1 | 3.2 | 3.6 |
| Excluded element present | 1/10 | 3/10 | 2/10 |
| "Most interesting" listener pick | 3 | 6 | 1 |
Mock interpretation: v6-wild loses adherence but wins on surprise, which supports the "scout with wild, finish with v6" workflow.
Follow-ups
- Repeat with Variety at Normal to measure how much prompt rewriting costs in adherence.
- Test whether raising Style Influence closes v6-wild's gap.
Sources
- Suno Help: What's new in v6?help.suno.com/en/articles/13924801
- HookGenius: Suno v6 Prompt Guidehookgenius.app/learn/suno-v6-guide/