What do the 'Weirdness' and 'Style Influence' settings do in Suno AI?
Weirdness and Style Influence are two sliders in Suno's Custom mode. Weirdness runs Safe to Chaos, and Suno documents 50% as the "normal" expected result — higher settings pick less predictable musical choices. Style Influence runs Loose to Strong, setting how literally Suno obeys the tags in your Style field.
Why — the first-principles explanation
Suno builds a song by repeatedly guessing what should come next — the next drum hit, the next chord, the next vocal phrase. At every one of those moments the model does not have a single answer. It has a ranked spread of possibilities, each with a probability attached: the obvious four-on-the-floor kick might sit at 60%, a syncopated one at 15%, something genuinely strange at 0.3%. A finished track is just thousands of these small picks chained together. Both sliders are steering that pick — they just steer different parts of it.
Weirdness decides how far down the list Suno is willing to reach. Turn it toward Safe and the model keeps grabbing near the top of the ranking, which produces music that sounds correct, familiar, and sometimes forgettable — you are hearing the most probable version of your genre. Turn it toward Chaos and low-probability options get picked far more often: odd instrument choices, transitions that do not resolve where your ear expects, rhythms that fight the grid. Neither end is "better." Safe risks blandness; Chaos risks a song that falls apart, because one unlikely pick early on forces the model to keep building on a shaky foundation. Suno's own documentation puts the neutral point at 50%.
Style Influence decides how hard your text pulls on that whole process. Your Style field — "dark synthwave, female vocals, 90 BPM" — is a condition applied to every single pick. Low Style Influence lets that condition act as a gentle suggestion, so the model drifts toward whatever it finds musically natural and your tags become flavor rather than law. High Style Influence forces every pick to stay inside the lane your tags describe. The catch is that strong conditioning narrows the model's options, and a narrowed model tends toward the generic and the literal — crank it to maximum and you can get a track that hits every tag while sounding like a stock-music demo of those tags.
This is why the two sliders are independent axes rather than one "creativity" dial, and why they can pull against each other. High Weirdness plus high Style Influence means be surprising, but only inside this box — often the most useful combination. High Weirdness plus low Style Influence removes both guardrails at once. One honest caveat: Suno has not published what these sliders map to internally. Their described behavior matches two well-understood knobs in generative audio research — sampling randomness and conditioning-guidance strength, where turning guidance up is known to trade diversity for obedience — but treat that as a strong analogy, not a confirmed spec.
An example that makes it click
Imagine you hire a 4-piece bar band and hand the singer a sticky note that reads: "sad 80s synth-pop." Style Influence is how strictly the band reads that note. At Loose, the guitarist glances at it, shrugs, and plays the blues song he likes — a little sad, sure, but not really 80s. At Strong, the band treats the note like a contract: synths on, tempo down, no blues allowed, every tag honored.
Weirdness is a separate question: how adventurous is the drummer allowed to be? At Safe, he plays the beat you have heard on 500 songs. It works. It is boring. At Chaos, he decides the chorus needs a cowbell in 7/4 — which is either the best part of the song or a train wreck, and you will not know until you press play. Notice the two dials do not overlap. You can absolutely have a strict-but-adventurous band (Strong note, weird drummer) — that is usually the good one. A loose note and a chaotic drummer just gives you four people playing different songs.
How to do it
- Open Suno and switch to Custom mode — the sliders do not appear in the simple one-box prompt view.
- Fill in your Style field first with real tags (genre, mood, instruments, tempo). The sliders do nothing useful until there is a style for them to act on.
- Open Advanced Options to reveal the Weirdness and Style Influence sliders.
- Leave Weirdness near 50% for your first generation — Suno documents this as the normal expected result, so it is your baseline for comparison.
- Diagnose what went wrong, then move ONE slider. Song ignored your tags? Raise Style Influence. Song obeyed your tags but sounded generic and lifeless? Raise Weirdness. Song fell apart into noise? Lower Weirdness.
- Regenerate and compare against your baseline. Moving both sliders at once tells you nothing about which one caused the change.
- If you uploaded an audio file, a third slider — Audio Influence — appears; it governs how closely the output tracks your uploaded audio rather than your text.
Key facts
- Suno's official documentation describes Weirdness as going from Safe to Chaos, and states that 50% is the "normal" expected result.
- Suno's official documentation describes Style Influence as letting you choose how close you stay to your style input, from Loose to Strong.
- A third slider, Audio Influence, appears only when you are using an Audio Upload.
- All three sliders are located in Custom mode; Suno files the help article under version V4.5 (as of 2026-07).
- Suno has not published which internal sampling parameters these sliders control — no official mapping to temperature or guidance scale exists as of 2026-07.
- In published generative-model research, classifier-free guidance strength (Ho & Salimans, 2022) creates a documented trade-off between sample fidelity/adherence and diversity — the same trade-off users report when pushing Style Influence to maximum.
Generate full songs — vocals and instruments — from a text prompt.
▶ The 60-second explainer (script)
Two sliders in Suno's Custom mode, two totally different jobs. Here's the difference in ninety seconds. Weirdness runs from Safe to Chaos. Suno's own docs say fifty percent is the normal expected result — that's your baseline. Style Influence runs from Loose to Strong, and it controls how literally Suno obeys the tags you typed in the Style field. Now, why are these separate? Because Suno writes music by guessing what comes next, thousands of times in a row. At every guess it has a ranked list of options. Weirdness decides how far down that list it's willing to reach — Safe grabs the obvious choice, Chaos grabs the strange one. Style Influence decides something else entirely: how hard your text pulls on every single one of those guesses. Loose, and your tags are a suggestion. Strong, and they're a contract. Here's how to actually use them. Song ignored your prompt? Raise Style Influence. Song followed your prompt but sounded like stock music? Raise Weirdness. Song fell apart into noise? Drop Weirdness back down. And move one slider at a time — change both and you'll never know which one did it. One honest note: Suno has never published what these sliders control under the hood. The behavior matches known research knobs, but that's an analogy, not a spec.
What authoritative sources say
People also ask
What is the best Weirdness setting?
There is no official "best" value — Suno only documents 50% as the normal expected result. Use 50% as your baseline, then move up if outputs feel generic and down if they fall apart.
Does maxing out Style Influence make songs match my prompt better?
It makes them match more literally, which is not always better. Very strong conditioning narrows the model's options and often produces a track that hits every tag while sounding like a stock demo of those tags.
Why did Suno ignore my style tags completely?
Low Style Influence treats your tags as loose suggestions rather than rules. Raise the slider toward Strong. Also check that your tags are real musical descriptors — vague or contradictory tags give the model nothing firm to obey.
What is the Audio Influence slider?
It is a third slider that appears only when you use an Audio Upload, per Suno's documentation. It governs how closely the generated track follows your uploaded audio instead of your text prompt.
Can I use these sliders without Custom mode?
No. Suno's documentation places the creative sliders in Custom mode. The simple single-prompt view does not expose them, so you have to switch to Custom to get access.