How to Make an AI Cover of a Song in Another Voice — or Your Own
AI re-sings the vocal line in a different voice, or in yours, while the melody, timing and arrangement stay put. Listen to two before-and-after examples and see how to make one.

Contents
Want to hear your song as a duet partner would sing it, pitch a demo in a different register, or just hear a favourite track in your own voice without booking a studio? An AI cover doesn't rewrite the song. The AI re-sings the vocal line note for note, and the melody, timing and arrangement stay exactly where they were. Only the voice changes.
Before and after
Our example is “Stay a Little Longer”, an acoustic ballad with a male lead written by MiniMax Music on ImageGPT. Here's the original, then the same song in two of the built-in voices.
How it works
First, Demucs separates the vocals from the backing track. Then YingMusic, a singing voice conversion model, re-sings the vocal line in the voice you picked — same notes, same rhythm. The new vocal is mixed back over the instrumental and you get a finished MP3.
Pick a voice
| Voice | Sound |
|---|---|
| Bright | Female, clear and sweet pop soprano |
| Alto | Female, warm and breathy — soul and jazz |
| Power | Female, strong rock belt with some grit |
| Tenor | Male, smooth and light pop voice |
| Baritone | Male, deep and warm crooner |
| Raspy | Male, husky rock voice |
All six are synthetic voices, and you can preview each one with the Listen button before you spend anything. There are no celebrity voices.
What people use it for
- A duet partner. You wrote a song for a male voice and want to hear how the female part would sound before you look for a singer — run it through Bright or Alto.
- A cleaner demo. A rough take sung over a guitar becomes a clean, steady baritone or tenor vocal you can send to bandmates.
- Your own voice on a favourite track. Upload a short clip of yourself and hear the song as if you'd recorded it — just keep that one for personal listening.
- Audio from a video. Drop in an MP4 of a performance and get back an MP3 cover, with artwork if you switch it on.
- A song you made with AI. If you generated a track with a vocal you don't love, swap the voice and keep the arrangement.
For the best result, start from a clean recording: a studio or streaming-quality file separates better than a live phone recording with crowd noise, and the cleaner the vocal stem, the cleaner the new voice.
Step by step
- Open AI Cover Generator under Audio and upload a song — audio or video, from 10 seconds to 6 minutes.
- Choose one of the six voices, or My voice. For your own, upload 8–60 seconds of voice with no music (singing works best, talking works too) and tick “This is my own voice, or I have this person's permission”.
- Set the mix: Vocal level (Quieter, Normal or Louder) and Reverb (None, Light or Hall).
- Optionally add a title and turn on Cover art: the AI paints square artwork from the song's lyrics and mood and embeds it in the MP3.
- Tick “I have the rights to this song, or the cover is for personal listening only” and click Make cover.
- Compare the result with the original right in the tool, then download the MP3.
Whose voice, whose song
Only clone your own voice or the voice of someone who has agreed to it — you confirm this before every run. As for songs: use your own, ones you hold the rights to, or anything at all if the cover is just for personal listening. Before posting a cover of someone else's song, get permission from whoever owns it.
Pricing
The price depends on the song's length and is shown before you start. If the cover fails, you get your credits back.
| Song | Credits |
|---|---|
| 3 minutes | 33 |
| 3 minutes with cover art | 40 |
| 4 minutes | 42 |
Upload a song, pick a voice and compare the cover with the original right in the tool.
Make an AI coverTry everything from this article in minutes
All the models and tools from this article are in one place. You get 6 free credits after sign-up, and every price is shown before you run it.
Make an AI cover

