Add an AI Instrumental to Your Vocals

Music made to fit a voice

Songwriters often start with the voice: a recorded melody on a phone, a rough topline, a finished acapella. Turning that into a band arrangement is where projects stall. This page generates the accompaniment for you, building the chords, bass and rhythm around the melody you upload.

The result lets you hear your vocal with real instruments around it in one pass, which is much faster than programming parts by hand to test an idea.

Describing the accompaniment

The tags field is your arrangement brief. Name the instruments you want to hear, the genre, and the energy: warm upright bass, brushed drums, soft piano, slow tempo. Specific instrument names pull the result in a clear direction more reliably than abstract mood words.

The negative tags field is the other half: use it to exclude sounds that would fight the vocal, such as heavy distortion, aggressive synth leads or sub-bass drops. Both fields accept up to 500 characters.

Recording a better source

Because the model has to work out the melody from your audio, a clean recording matters. Sing close to the microphone in a quiet room, keep reverb off, and avoid background music. A steady tempo helps the generated rhythm lock onto the phrasing.

If the first result misses the feel, vary the tags first and the upload second. A calmer take with clearer pitch tends to improve the harmony the model chooses beneath it.

Limits worth knowing

The page accepts a normal audio upload, so it does not need a track generated earlier on the site. The model infers harmony from the melody you sing, so pitchy or very reverberant takes can lead to unexpected chords. Title, style tags and negative tags are all required. Results vary between runs, so generate a few versions at 9 credits each and compare them. A clean, dry source recorded close to the microphone gives the model the clearest pitch to follow, which in turn produces more sensible chord choices.