What is an AI song cover?
An AI cover gives an existing performance a different singing voice. The source supplies the words, melody, rhythm and phrasing; your trained voice supplies the vocal character, alongside the song's accompaniment.
Use a permitted song with an available official voice or a voice from your private library. Hear a different singer on a finished arrangement, compare voices for a demo, or revisit your own song.
From a song to a new vocal version
Upload your source song
Choose a recording with the singing and arrangement you want to keep and confirm your upload rights. Upload MP3, WAV, M4A or FLAC, up to 25 MB and 10 minutes.
Choose or train a singing voice
Select a ready voice. For a new voice, upload or record 2–30 minutes of clean solo singing and review its training price. Once trained, you can reuse it across songs.
Review the price and create
Choose the original pitch or a vocal pitch shift and review the quote. Lacuna separates, converts and remixes the vocals with the accompaniment. Listen to the result in your workspace.
A new singing voice, a familiar arrangement
Keep the accompaniment with your new vocal
Choose official voices or build your private library
Save a new track and compare your versions
Hear the same song through a different voice
Explore the singer and song together.
Put your voice on a songwriting demo
Use your trained voice on a recorded guide vocal. Hear how it fits the melody and arrangement before planning another recording session.
Compare voices for a collaboration
With each singer's permission, compare their trained voices on the same song, including verses, choruses and sustained notes.
Revisit a song you have already made
Keep your song's lyrics and backing track, try a ready voice, and save another vocal version beside the original.
Choose material that gives your voice room to work
Prepare the source performance and training recording with these checks.
Start with a clear lead vocal
Choose an audible main singer. The cover follows their pronunciation, note lengths and phrasing. An instrumental has no singing performance to convert.
Train with clean solo singing
Record one singer without backing music for 2–30 minutes. Aim for 10–20 minutes covering low, middle and high notes, plus quiet and stronger phrases.
Match the song to the voice's range
Begin at the original pitch. Shift the converted vocals by up to 12 semitones in either direction. The accompaniment stays in its original key.
Prefer fewer overlapping voices and effects
Harmonies, choirs, reverb and heavily processed singing can make separation harder. A simpler source helps you judge the selected voice.
Check the song's format and length
Songs must fit within 25 MB and 10 minutes. If unreadable, export a fresh MP3, WAV, M4A or FLAC. Training recordings require 2–30 minutes.
Listen beyond the first few seconds
Compare verses, choruses, high notes and quiet endings. Listen for clear words and natural transitions, then judge the voice's fit.
Making AI song covers
Choose an available official voice, or train your own voice from 2–30 minutes of clean solo singing. Upload a song and select a ready voice. Training and conversion each show their price before submission.
It follows the source's words, melody and timing, with optional vocal pitch shift. To compose from text, use AI Song Generator with My Voice.
It converts singing already present in audio. Lyrics or an instrumental alone do not supply that performance. Use the song generator to start with text.
Upload MP3, WAV, M4A, FLAC or another supported audio file, up to 25 MB and 10 minutes, with an audible lead singer.
No. Lacuna separates, converts and remixes the vocals with the accompaniment. Suitable stems already available for the same source track can be reused.
No. Available official voices and ready voices from your private library can be reused across songs and sessions without training another model.
Select a ready voice to see conversion and required separation costs. Reusing suitable stems avoids another separation charge. New voice training has a separate quote.
Yes. Shift the converted vocals between −12 and +12 semitones; the accompaniment keeps its original pitch. This does not automatically correct individual notes.
Play the separate result track in your workspace. The finished mix is MP3. Workspace audio downloads require a paid plan.
Use permitted songs and another singer’s recordings only with explicit permission. Official voices are shared library options; voices you train stay private. Creating a cover grants no rights to the source recording, composition, lyrics or another person’s voice.
Build your voice and your next song
Train a reusable singing voice and manage names and avatars in your private library.
Create an original song from a description or lyrics with a trained voice from your library.
Get individual vocal and instrumental stems for listening, editing and your next music workflow.
Hear your singing voice on the next song
Choose a recording and a ready voice, or prepare your first singing voice. Review the quote before creating your cover.