You switched vocals off. Someone is still humming.

It is one of the most common complaints about Suno. You ask for an instrumental, and the render arrives with a wordless "ooh" over the chorus, a chopped vocal sample in the drop, or a full verse in a language nobody asked for.

Suno is not ignoring you. It is following other parts of your prompt that point the other way. Most genres are learned from songs with vocals, so the model's default for "a track in this genre" includes a voice. An instrumental has to be defended, in several places at once, or the default creeps back.

This guide covers the four places that matter, a check you can run on any prompt in ten seconds, and three full instrumental templates at 1,000 characters each.

Where the vocals come from

When an "instrumental" comes back with a voice, it is almost always one of these:

The lyrics box had text in it. Even a single line gives Suno something

to sing. Section tags on their own are fine; words are not.

The genre implies a singer. Pop, R&B, gospel, reggaetón and most dance

sub-genres are overwhelmingly vocal in the material Suno learned from. The genre word carries a voice with it.

A voice word hides elsewhere in the prompt. Anthem, hook, chant,

choir, vocal chop, vocal-forward, MC, rap. Any of these, in any field, is an instruction to put a voice in the mix.

Nothing says no. If the prompt never mentions vocals at all, the genre

default decides.

The first and last are easy to fix. The third is the one that catches experienced users, because the word is doing a different job where it sits.

We found the third problem in our own generator

When we pulled the templates for this article from GENPROMPT's generator, we checked every candidate for voice words outside the vocals field. On 27 Sep 2026 we sampled 4,000 prompts from each of three modes and kept only the ones whose vocals field already said no vocals:

Amapiano: 164 of 375 instrumental prompts (44%) still described the style

as vocal-forward somewhere else in the prompt.

UK drill: 0 of 367.
Drift phonk: 0 of 483.

So in almost half of the amapiano cases, the prompt said "no vocals" in one field and "vocal-forward" in another. Suno has to pick one, and the genre's own default is on the side of the voice. That is a bug in our own generator, and we would rather publish the number than hide it. The lesson applies to any prompt you write by hand: one "no vocals" does not cancel a "vocal-forward" three lines away. It just makes the prompt argue with itself.

The four switches

1. Suno's Instrumental toggle

In custom mode, switch Instrumental on. It removes the lyrics text, which removes the most direct source of a voice. It does not stop the model adding wordless vocals or vocal samples on its own, which is why the other three switches exist.

2. The vocals field says no

Write the vocals value explicitly: no vocals, instrumental, followed by what takes the voice's place. No vocals, pure instrumental Memphis drift loop tells Suno both what is missing and what fills the space. An empty field says nothing.

3. The voice-word sweep

Before you create, search the prompt for these words and remove or replace each one outside the vocals and negative fields:

vocal, vocals, vocal-forward, vocal chop, hook, chant, choir, rap, MC, shout, sing, singalong, anthem

Anthem is the one people forget. It rarely appears as an instruction, but in the material Suno learned from, anthems are sung.

4. The Exclude field

If your Suno plan shows the Exclude field, paste this into it:

vocals, singing, humming, choir, vocal chops, rap, spoken word

Exclude blocks these harder than a "no" inside the style prompt. The negative prompts guide explains the difference between the two layers and why using both is stronger than either alone.

Three full instrumental templates

These three were generated by GENPROMPT's own modes on 27 Sep 2026 and are pasted unedited. Each has no vocals in the vocals field, passed the voice-word sweep above, and is exactly 1,000 characters long, the full size of Suno's style box.

Template 1: drift phonk, 142 BPM

The cleanest of the three: no vocals appears twice, in the vocals field and in the negative field, and the cowbell fills the space a voice would take.

{"style":"viking phonk Norse aesthetic dark 2024-2025 2026, [142BPM C# minor + Turkey drift phonk culture TikTok + 2024-2025 TR]",
"length":"1:30 minutes","bpm":142,"drop":"bar 4 beat 1","key":"C# minor",
"kick":"aggressive Memphis kick saturated 200-800Hz",
"bass":"detuned 808 dark, pitched -3st, drift phonk signature",
"melody":"cowbell sample chopped rhythmic Memphis signature B2-F#3",
"vocals":"no vocals, pure instrumental Memphis drift loop",
"percussion":"gritty cowbell 8th + open hat offbeat + rimshot dry",
"swing":"mechanical no swing 25% drift tight grid forward",
"structure":"no intro, full bar 1, 98% loop",
"texture":"cold digital dark, distortion heavy, no warmth",
"atmosphere":"car engine rev texture distant",
"production":"kick and sub bone-dry upfront, FX only on lead",
"era":"2023-2025 phonk drift mainstream era",
"dynamic":"plateau 90% phonk constant menacing",
"mood":"Memphis dark underground threatening",
"negative":"no vocals, no funk, no swing, no bright elements"}

Template 2: amapiano, 116 BPM

A pitch-shifted log drum as the lead, jazzy warmth in the style line, and no voice words anywhere else. This template carries no negative about vocals, so the Exclude list above matters more here.

{
"style":"amapiano origin Pretoria SA 2012-2014 log drum invention township raw foundation Zulu, amapiano origin Soweto 2013 [116BPM G major + SA deep house pre-amapiano transition log drum jazzy + 2013-2015 SA]",
"length":"3:30",
"bpm":116,
"drop":"bar 16 beat 1",
"key":"G major",
"kick":"amapiano kick punchy tight 60Hz body warm sub",
"bass":"sub bass amapiano D1 deep underground",
"melody":"log drum pitch-shifted octave amapiano 2024 modern",
"vocals":"no vocals instrumental amapiano groove SA cut",
"percussion":"log drum rolls amapiano signature off-beat SA",
"swing":"medium swing 28% amapiano female-DJ",
"structure":"plateau 95% amapiano constant pocket",
"texture":"warm SA mids log drum sustained tonal",
"atmosphere":"log drum echo space amapiano room verb",
"production":"2018-2025 Johannesburg authentic",
"era":"2018-2025",
"dynamic":"build-drop amapiano log drum into peak",
"mood":"deep amapiano underground reflective soulful",
"negative":"no trap structure, no 808 dominant"
}

Template 3: UK drill, 141 BPM

Sliding 808 as the melody, cold atmosphere, and no hook anywhere: no pop hooks in the negative field works as a second, indirect vocal block.

{
"style":"UK drill melodic crossover 2022-2025, Brooklyn drill 2018 NYC dark [141BPM C# minor + menacing UK production exported sliding 808 + 2018-2020 NYC]",
"length":"3:45",
"bpm":141,
"drop":"bar 4 beat 1",
"key":"C# minor",
"kick":"aggressive UK drill kick distorted saturated",
"bass":"sliding 808 portamento drill UK trademark",
"melody":"sliding 808 detuned dark London type signature",
"vocals":"no vocals instrumental dark drill atmospheric UK",
"percussion":"open hat staircase drill rolls UK signature 141BPM",
"swing":"tight UK drill grid mechanical 0% swing precision",
"structure":"pulse-wave 808 sliding pattern menacing roll",
"texture":"cold dark UK drill mids massive sub heavy plate",
"atmosphere":"dark UK rain reverb cinematic 141BPM bed",
"production":"2019-2020 Brooklyn drill crossover",
"era":"2019-2020",
"dynamic":"pulse-wave 808 sliding pattern menacing roll",
"mood":"dark UK street menacing London 2022 cold",
"negative":"no major key, no bright melody, no pop hooks"
}

What to put in the lyrics box

With Instrumental on, nothing. If you need the lyrics box for structure, for example to force a break, use section tags only and describe instruments in parentheses, never words to be sung:

[Intro]
(pads and log drum only)
[Drop]
[Break]
(piano alone, no drums)
[Drop]
[Outro]

If a render still comes back with a voice, check the style for hook or anthem before regenerating. Rewriting the prompt usually fixes it faster than rolling the dice again.

Turning a vocal prompt into an instrumental one

You do not need a new prompt to get an instrumental version of a track you like. Change three things and keep the rest:

Replace the vocals value with no vocals, instrumental plus the

instrument that takes the lead.

Run the voice-word sweep on the style, structure and mood.
Put the Exclude list above into Exclude.

Keep the total at 1,000 characters or less. Suno cuts anything past that without warning, and the end of the prompt, where the negative field sits, is the first thing to go.

Generate your own

The widget at the top of this page generates a fresh, full-length prompt with every click, free and without an account. For instrumentals, pick a genre, look at the vocals field, and apply the three changes above if it asks for a voice.

Suno Instrumental Prompts: How to Stop the Surprise Vocals