Short setup
Between fifteen and thirty seconds of singing is enough to create a model, recorded in the browser or uploaded.
Clone My Voice
Build a voice model from your own singing and use it on songs you have rights to.
1
2
3
4
Sing for 15 to 30 seconds, or upload a recording of your own voice.
For the best model, record one voice alone, with no music behind it, for 15 to 30 seconds.
or
My Voices
Sign in to see your cloned voices.
your voice as an instrument
What makes a good voice sample
Record a short clean vocal, create a personal voice model, and use it to hear songs sung the way you would sing them.
The model learns from whatever it hears, including room echo and background sound. A dry recording of one voice with no music behind it produces the most faithful result.
You can record directly in the browser or upload a file. If the take is longer than needed, choose the strongest thirty-second section so the model hears your most typical tone.
Voice cloning on SeedAudio is designed for one person: you. Sing or upload 15 to 30 seconds of your own unaccompanied voice, and the tool builds a model you can reuse across songs.
The model is useful for demos, for trying keys and styles you have not recorded yet, and for covers of songs you have rights to. It is not a way to reproduce other people's voices.
How the voice model is used
Reusable across songs
Once created, the voice is available for covers and new songs, so one sample serves many projects.
Verification built in
You record a short verification phrase with your sample, which ties the model to the person actually singing.
From recording to model
Find a quiet room and warm up first. The sample is short, so the quality of those seconds matters.
- 1
Record or upload
Sing a comfortable phrase for at least fifteen seconds with no music playing, or upload an existing solo vocal.
- 2
Choose the best section
Trim to the strongest thirty seconds so the model hears a steady, representative tone.
- 3
Create and use the voice
Build the model, then pick it when making a cover of a song you are allowed to use.
Why singers clone their own voice
Demo without a session
Hear a new song in your voice before booking studio time or rehearsing it properly.
Testing keys and styles
Try a genre or register you have never recorded to decide whether it suits you.
Writers who do not sing
A songwriter can license a singer's model with their permission to produce convincing demos.
Sample recording tips
One voice, no backing
Harmonies, instruments or a playing speaker confuse the model. Record your voice alone.
Close and dry
Stay close to the microphone in a soft-furnished room to keep echo out of the sample.
Sing naturally
Use your usual tone rather than extreme techniques; the model generalises best from typical singing.
Voice clone questions
How long does the voice sample need to be?
At least fifteen seconds and up to thirty seconds of clean solo singing. Longer uploads can be trimmed to the best thirty-second section.
Can I clone a celebrity or another singer?
No. Only clone your own voice, or a voice whose owner has given you explicit permission.
What can I do with the voice model?
Use it for voice covers of songs you have rights to, or to sing new songs you generate.
Why does my model sound different from me?
Background noise, reverb or music in the sample changes what the model learns. Re-record in a quieter space with no backing.
Build a voice model from your own singing
Record half a minute of clean vocals and hear your next song in your own voice.
Clone my voice