What an audio input actually is
An audio input is a sound you give Suno to work from instead of - or alongside - a text prompt. You record it on the spot or upload a file, and from that moment it lives in your Library as a source you can extend, cover, remix, turn into a Persona or drop onto a Studio timeline.
The distinction that matters: a prompt describes a song that doesn't exist yet. An audio input hands over something real - a melody, a groove, a voice, a room - and asks Suno to build around it. That's a different creative act, and it's why the feature changes how the whole tool feels.
What a clip can become
Once the audio is in, Suno offers four fundamentally different things to do with it. Picking the wrong one is the most common reason people conclude the feature "doesn't work" - they wanted a cover and asked for an extend.
ExtendKeep the audio, continue it
Your clip stays exactly as it is and Suno writes what comes next. Set a timestamp to continue from, give it a style, add lyrics if you want them. The original recording is still your recording - it's the front of the song.
This is the route for a hummed idea, a half-finished verse, or a riff you never turned into anything.
CoverKeep the song, replace the sound
Suno keeps the melody and structure and re-performs the whole thing in a style you describe. Your original audio doesn't appear in the output - it's the blueprint, not the building.
A phone voice memo of you singing becomes a full band arrangement. A piano sketch becomes a synthwave track.
Persona or VoiceKeep the identity, reuse it
Turn the clip into something reusable. A Persona captures the vocal character, genre and production texture so later songs inherit it. Voices - live on Android and iOS since 7 August 2026 - records your own voice once so it can sing anything you write.
This is how you get a consistent artist identity across a catalogue instead of a different singer every generation.
Studio timelineKeep everything, arrange it yourself
In Studio 2.0 the clip goes straight onto a track and you build around it - AI parts on other lanes, your recording untouched on its own. Studio also records directly, so a vocal or a guitar take can go in without ever becoming a file.
Premier only, desktop browser only, and the only route where your audio survives to the export at full quality.
Audio Influence
Under Advanced Options, once a clip is attached, sits the one slider that decides whether this feature works for you. It isn't a quality setting - it's a relationship setting. It answers a single question: how strongly should the result stick to the audio I gave you?
Low and your clip is a vague suggestion the model can ignore. High and the melody survives but the arrangement has less room to become anything new. Around 55% is where most people land: the contour and timing come through, everything around them gets fully re-rendered.
These two pull against each other. Push Audio Influence up and your style tags get less say; push Style Influence up and your recording gets reinterpreted harder. Move one at a time. Changing both and disliking the result tells you nothing about which one caused it.
Set it by what the clip is for
Before you touch the slider, decide what job the audio is doing. That single decision determines the setting, and it's the step most people skip.
| Your clip is a… | Audio Influence | Because |
|---|---|---|
| Lead vocalA take you want to hear in the output | High70-85% | Phrasing and timing are the performance. Use Extend or Studio, not Cover. |
| Melody guideYou humming the tune | Medium-high55-70% | The contour is the idea; the voice delivering it isn't meant to survive. |
| Riff or chord anchorGuitar, piano, a loop | Medium50-65% | You want the harmony kept and the rest of the band invented. |
| Groove referenceA tapped or beatboxed rhythm | Medium-low35-55% | Feel and tempo transfer well; the actual sound of your tapping shouldn't. |
| Texture or atmosphereRain, a room, a field recording | Low20-40% | You're setting a mood, not dictating notes. High settings turn ambience into structure. |
| Full demoA finished-ish song of yours | High70-90% | You're asking for production, not composition. Say so in the style prompt too. |
Preparing the clip
Suno is good at interpretation and bad at guessing what you meant to leave out. Two minutes of trimming beats twenty minutes of regenerating.
Trim the dead air
Cut false starts, count-ins and the four seconds of you finding the record button. Silence at the front reads as an intro and the model will happily build one there.
Go dry, not roomy
Reverb and heavy processing bake into the analysis and come back as metallic, smeared vocals. A close, dry, slightly boring recording produces a far better result than an atmospheric one.
Keep the tempo honest
Timing drift is the hardest thing for Suno to reconcile with a grid. If you're humming, tap your foot. If you know the BPM, put the number in the style prompt as well.
Workflows that actually work
Concrete combinations of route, slider and prompt for the four things people most often want from this feature.
Hum a tune, get a song
Record 20-40 seconds of you humming the melody, dry and in time. Choose Cover, write a full five-part style prompt, set Audio Influence around 60%.
You'll get the tune back arranged, performed and produced - your voice nowhere in it. If the melody drifts, the problem is the humming, not the setting.
Sing the verse, keep the take
Record the vocal properly - close mic, no reverb, headphones so there's no bleed. Choose Extend, set a timestamp at the end of your phrase, push Audio Influence to 75-85% and describe the band you want behind it.
Say in the style prompt that the vocal is featured and ask for space around it, or the arrangement will bury you.
Play a riff, build a band
Record eight bars of the part, looped cleanly. Upload it, choose Extend or drop it on a Studio track, and set Audio Influence around 55-65%.
Name the instrument in your prompt - clean jangly Telecaster, not guitar - so the AI parts arrange around what you actually played rather than doubling it.
Upload a finished song and take it apart
On Pro or Premier you can bring in a full track up to eight minutes. From there the Song Editor lets you reorder and rewrite sections, and stem separation splits it into as many as 12 parts.
This is the route for remastering something old of yours, pulling a usable vocal out of a rough mix, or rebuilding an arrangement from scratch around parts you already own.
What you're allowed to upload
Audio input is the one Suno feature where you can create a legal problem for yourself in a single click, because you're the one supplying the material. The rules are short and the enforcement is partly automatic.
Use material you created, control, or have written permission to use. The upload flow asks you to confirm exactly that, and ticking it is a representation you're making - not a formality.
Your own singing, your own playing, your own field recordings and your own old demos are all fine. A stem you pulled off someone else's release is not, whatever you plan to do to it.
Suno screens uploads and refuses ones it identifies as commercially released material. It isn't perfect in either direction - it will occasionally reject something you genuinely own that resembles a known track, and detection is not the same thing as permission.
If your own recording is rejected, that's the screen being cautious rather than an accusation.
Suno keeps uploads containing vocals private and unsearchable by default. It's a sensible protection - your voice memo doesn't become part of a public browsing surface - and it means an audio input isn't a publishing act.
Since the 3 September 2026 terms update, commercial rights attach to an official download obtained through Suno while on a paid plan - and the rights you hold in the output can't exceed the rights you held in the input. Feeding in audio you don't own doesn't turn the result into something you can release. Detail on the downloader page.
Fixing bad results
Almost every complaint about audio inputs maps onto one of these six, and five of them are fixed by changing one thing rather than regenerating hopefully.
sparse arrangement, guitar forward, drums stay out of the verse. Or move to Studio, where you control the lanes.
Audio input FAQ
Can I upload audio on a free account?
Yes, up to 60 seconds. Pro and Premier raise that to 8 minutes, which is what makes uploading a complete song possible. The feature launched as paid-only and has since opened up.
Where is the upload button?
On the Create page, directly next to the Create button, and again in your Library. In the mobile apps it's the row under the prompt box - hum, tap or upload.
Will my own voice be in the finished song?
With Extend or a Studio timeline, yes - your recording is kept and built around. With Cover, no: Suno keeps the melody and re-performs everything, so the voice you hear is generated.
What should I set Audio Influence to?
Start at 55%. Raise it toward 80% when the melody or performance must survive; drop it toward 30% when the clip is only setting a mood. Change it on its own, not alongside the style prompt.
Can I upload a video?
Yes - Suno pulls the audio track out of it. A clip filmed on a phone is a perfectly good way to capture an idea if that's what you had running.
Do uploads use credits?
Uploading itself doesn't. Generating from the clip does, exactly like any other song, and stem separation costs extra on top. Store as many source clips as you like at no cost.
Can I upload a song by another artist?
No. Suno blocks recordings it identifies as commercially released, and the upload flow asks you to confirm you own or have permission for the material. Whatever comes out the other end wouldn't be yours to release anyway.
Is my upload public?
Uploads containing vocals are kept private and unsearchable. Publishing anything you make from them is a separate, deliberate step.
What's the difference between an audio input and Voices?
An audio input is one clip for one project. Voices captures your singing voice as a reusable identity you can apply to any future song, and it needs a short verification recording. Different jobs, both starting from a microphone.
Can I get stems from an uploaded song?
Yes, on a paid plan. Upload the track, then run stem separation - you get up to 12 time-aligned parts back in your Library as individual items. Download them as WAV if they're headed for a DAW.
Does the clip stay in my Library?
Yes. Uploads persist and can seed new projects whenever you want - upload the riff once and reuse it for months.
Can I do all this on my phone?
Recording, uploading, extending and covers, yes. Studio 2.0 - the timeline route - won't run on a phone: it needs a 768px minimum screen width. More on the Android app page.
The best prompt you'll write this week is a hum
Thirty seconds of a melody you can't get out of your head will get you further than any amount of adjective-stacking. Record it badly, upload it, and let Suno do the arranging.