Remix and upload your own audio in Suno: turn a hum into a track
Upload a hum or a clip as a seed and let the Style prompt build a finished track around it: how to steer the arrangement, keep key and tempo coherent, extend from an upload, and the copyright rule that never changes.
An upload gives Suno the melody; the Style prompt decides everything built around it. When you hum a tune into your phone or drop in a rough clip, Suno hears pitch, timing, and roughly what key you landed in, then treats that as a seed. What comes back, the instruments, the genre, the drum feel, the vocal, comes almost entirely from the words you type in the Style field. Get the seed and the Style pulling the same direction and a shaky 20-second hum turns into a full track. Get them fighting and Suno hands you something that technically used your audio and sounds like neither the hum nor the genre you asked for.
What the upload actually does, and what it does not
Think of the upload as the skeleton and the Style prompt as the flesh. The clip carries the tune and the harmonic center; it does not carry taste. Suno will not guess that your phone-memo hum was meant to be a torch-song ballad or a drill beat. If you upload and leave the Style empty, you get Suno's most average reading of whatever it heard, usually a mid-tempo pop-ish arrangement that sands the character off your idea.
So the mental model is simple: the upload answers "what notes and when," the Style answers "played by whom, on what, in what style, with what voice." A hummed four-bar loop plus warm indie folk, fingerpicked acoustic, brushed drums, intimate male vocal is a completely different song from the same hum plus dark synthwave, analog bass, gated reverb drums, cold female vocal, even though the melody underneath is identical. That is the whole point of uploading instead of generating from scratch: you keep the one thing that was yours, the tune, and rent everything else.
Steer the upload with the Style field
The words that move the needle most are the same ones that matter in any Suno prompt: genre, instrumentation, vocal character, and production texture. The difference with an upload is that you are describing an arrangement to wrap around a melody that already exists, not conjuring a melody from adjectives. That changes how you write.
Lead with the genre and the core instrument, because those set the whole frame. Then name the vocal, because Suno will otherwise default to a generic delivery that ignores the emotion in your hum. Then add two or three texture tags to fix the mood. Resist the urge to stack ten descriptors; on an upload, a shorter Style follows the seed more faithfully, and a bloated one starts overriding the melody you were trying to keep.
Here is a worked example. Say you hummed a slow, wistful four-bar melody and want it built out as a late-night acoustic song.
Style:
warm indie folk, fingerpicked steel-string acoustic guitar, upright bass, brushed snare, intimate breathy male vocal, room reverb, analog warmth, slow ballad feel
Lyrics (the first line is the vocal anchor, not sung; it tells Suno who is singing):
[Intro]
(close-mic male vocal, soft and unhurried, no belting)
[Verse]
The porch light hums the only tune out here
I kept your record spinning through the year
[Chorus]
So play it low, play it slow
Let the needle carry what I never told
[Outro | fingerpicked guitar fades under room noise]
Exclude:
heavy drums, EDM synths, autotune, key change, double-time, aggressive vocal, brass
The Exclude field is doing quiet but real work here. key change and double-time are the two things Suno most often sneaks in when it decides your upload is "too simple" and tries to add drama. Naming them as things to keep out is how you tell Suno the plainness is deliberate.
Keep the key and tempo coherent
The single most common way an upload goes wrong: the Style prompt describes music in a different key or a different pulse than the clip. Suno then has to choose, and it usually splits the difference into something muddy. Your hum is your ground truth. Do not fight it.
For key: you rarely know the exact key of your own hum, and you do not need to. What you need to avoid is Style words that imply a fixed key or a modal color that clashes. Tags like dark minor, bright major, or phrygian will pull against a hum that already has its own tonal center. Leave the harmony language soft (warm, wistful, hopeful) and let the upload set the actual notes. If you do want to move the harmony, do it by re-humming, not by ordering a key change in text, which Suno applies as an abrupt jump rather than a re-harmonization.
For tempo: describe feel, not numbers. A hum has a built-in pulse, and a hard 128 BPM tag can collide with it. Words like slow ballad feel, mid-tempo groove, or driving steer the energy without picking a fight with the timing already in your clip. If the returned track feels like it is dragging against your melody, the fix is usually to soften or remove the tempo word, not to add a more specific one.
Extend and continue from an upload
An upload is not only a one-shot seed. The stronger workflow is iterative: seed a short clip, generate, then continue from the best result to grow it into a full arrangement. You are building a song in passes instead of demanding a finished one from a 15-second hum.
A reliable loop looks like this. Upload your hum and generate with a tight Style. Listen to the returned clips and pick the one whose feel is closest, even if it is short or imperfect. Then continue from that clip, and here is the key move: for the continuation, keep the same genre and instrument words so the section joins seamlessly, and only change the Style to signal what should shift, a [Chorus] lift, a bridge that strips back to just guitar, an [Outro] that fades. Each pass inherits the key and tempo from the audio before it, so coherence takes care of itself as long as you do not rewrite the whole Style between passes.
Two things to hold onto across passes. First, drift compounds: every continuation is generated from the previous audio, so a small wrong turn in pass two is baked into pass five. If a section comes back wrong, regenerate that section rather than continuing past it. Second, the vocal anchor line is worth repeating in each pass's Lyrics; without it Suno sometimes swaps the singer's character halfway through a song, and a voice that changes timbre at the bridge is the fastest way to make a track sound stitched together.
The rights caution, and why it outlives every UI change
This part does not change when Suno redesigns a button: what you upload has to be yours to upload. If you drop in a snippet of a commercial record, a stem you pulled from someone else's song, or a melody you lifted note-for-note from a track you do not own, you have not made an original song, you have made a derivative of copyrighted material, and no amount of restyling launders that. The output can still infringe even if it sounds different, because the seed carried protected melody and structure into it.
What is safe to seed with: your own humming, your own playing, audio you recorded, or material that is genuinely public domain or licensed to you for this use. What is not safe: ripped audio, a friend's unreleased track without permission, or a recognizable hook from a famous song. The practical harm is real and concrete. Platforms run content-ID matching, a track built on someone else's hook can get pulled, demonetized, or hit with a claim, and if you were planning to release or monetize, you have built the whole thing on sand. When in doubt, hum it yourself. A rough original hum you own beats a pristine sample you do not.
Common mistakes
Fighting the source key with your Style words. You hum something in a minor, wistful color, then type bright, uplifting, major-key pop. Suno cannot honor both, so it compromises into something off. Match the mood of your Style to the mood already in the hum, or re-hum first.
Uploading a clip that is too long. A 3-minute upload does not give Suno more to work with, it gives it more to average and more places to drift. A short, clean seed of a single strong idea steers far better than a long meandering one. Upload the hook, not the whole demo.
Leaving the Style empty and hoping. An empty Style hands your idea to Suno's default taste, which is exactly the generic sound you were trying to escape by uploading something personal. Always describe the arrangement you want around the tune.
A noisy or clipping recording. Suno hears the noise as musical information. Fan hum, table taps, and a distorted phone mic all leak into the pitch detection. Record the hum somewhere quiet and do not shout into the mic; a calm, clear hum reads far cleaner than a loud muddy one.
Rewriting the entire Style on every continuation. Change only what should shift between sections. If you swap the genre words mid-song, the continuation stops matching the audio it grew from, and the seams show.
FAQ
Can Suno turn a hum with no words into a full song with lyrics? Yes. The hum seeds the melody and the vocal follows it; you supply the words in the Lyrics field and a vocal anchor line so Suno knows who is singing. The tune stays yours, the words and arrangement come from the prompt.
Does the upload's audio quality end up in the final track? The character can, especially with a very noisy clip, because Suno reads texture as well as pitch. It is not sampling your recording into the master, but a clean seed gives a cleaner result. Treat the upload as a reference to be reinterpreted, not as a stem to be polished.
How do I keep the same voice across a long song built from an upload? Repeat the vocal anchor line in each continuation's Lyrics and keep the vocal descriptor identical across passes. The most common cause of a voice changing mid-track is dropping that anchor when you continue.
What if the generated track ignored my melody? Usually the Style was too heavy or clashed with the seed's key. Shorten the Style, drop any hard key or BPM tags, remove aggressive genre words that imply a different tune, and regenerate. A lighter prompt lets the upload lead.
Is it legal to upload a song I did not write and restyle it? No, not if it is copyrighted and you do not have the rights. Restyling a protected melody still produces a derivative work. Seed with audio you own or that is genuinely free to use. Check current terms and licensing before releasing anything built on an upload.