Create a singing voice you can use again
Upload or record solo singing, train a private voice, and give it a name and avatar. Use it for song covers and original songs across your music projects.
Hear two trained voices sing the same chorus
Julian and Stella are voice models trained on Lacuna. Compare each one with the original chorus of Almost Didn't Come.
Julian and Stella are trained voice models. Choose one and compare it with the original chorus.
Your singing voices, in one place
Sign in to add recordings, train a singing voice, and manage the voices saved to your account.
Voice creation and song conversion are charged separately. Review the quoted creation price before training; conversion prices appear when you use a trained voice.
A saved singer for the songs you want to make
AI singing voice cloning trains a reusable voice from recordings of one singer. Your recordings become the reference for that voice's sound, and the trained voice stays in your private library for future projects.
When you use it in a song, voice conversion changes the vocal timbre while following the source singing's melody, words and phrasing. Start with an existing song for a cover, or create an original song first and apply your saved voice to its vocals.
From your recordings to a ready singing voice
Add solo singing
Upload existing recordings or sing into your microphone. Use 2–30 minutes of clean singing from one person; 10–20 minutes gives you room for different notes, vowels and phrases. For a recording with backing music, turn on Extract vocals. Listen to each recording before saving it.
Review the quote and train
Give the voice a recognizable name and avatar, review its separate training price, then start training. Saving a recording does not start training automatically. You can leave the page and return to your library to check its status.
Choose your ready voice
Once training is complete, select the voice for an AI cover or an original-song project. Keep using the same saved voice across songs, with a separate conversion price each time. Listen to the result and consider the singer's natural range.
One singing voice, more ways to make music
A private voice library you can reuse
Give an existing song your vocal sound
Bring your voice to an original song
Choose the workflow that fits your next song
A reusable singing voice lets you return to the same singer while trying different songs, lyrics and musical directions.
For cover singers
Train your own singing voice, then try it on song recordings you have permission to use. Hear a different vocal sound while keeping the source song's melody and words. Reuse the same voice for another cover without training it again.
For original songwriters
Bring a lyric draft or a song idea and choose your saved singer. Develop the music in the workspace, then hear the generated song with your trained voice. Keep both versions available as you decide what to refine next.
For demo producers
Create a reusable voice from a singer you have permission to train and use. Apply it to your demo songs to explore how that vocal sound fits different musical directions. Compare the converted version with the source performance before choosing your next edit.
Prepare recordings for singing voice cloning
The recording gives training its reference. Spend a little time choosing clear solo singing with a useful variety of notes and phrases before you start.
One singer throughout
Use the same singer throughout the recording so the training material consistently represents one voice. Leave out duets, harmonies, backing singers and conversations. A clear solo performance gives you a more useful starting point than a mix of several people singing together.
Extract vocals from recordings with music
Dry solo singing gives training the cleanest reference. If a recording has backing music, reverb or other sounds, turn on Extract vocals for it, and the voice is isolated before training. Extraction works on recordings up to 10 minutes long and is charged per minute of that recording. Choose passages where the lead singer remains clear and other voices do not dominate.
Keep the recording clear and dry
Record in a quiet room and keep the microphone at a steady distance. Prefer a vocal with little background noise, room echo or added reverb. Avoid strong effects that obscure the singer's sound, and listen back for distortion or distracting noises before saving the recording.
Cover your natural singing range
Include comfortable low, middle and high notes rather than repeating one pitch for the entire recording. Use gentle and stronger phrases where they feel natural. Stay within the singer's usual range; forcing unfamiliar notes does not make the training material more representative of their voice.
Include different vowels and phrases
Mix held vowels with sung words, shorter phrases and longer lines. Familiar songs can help you collect different ways the singer forms notes and words. Aim for useful variety instead of filling the duration with long silence, spoken sections or the same noisy phrase repeated many times.
Prepare the complete recording
Upload or record 2 to 30 minutes in total, with 10–20 minutes recommended. You can add several recordings to one voice. Prepare any edits before adding a file, then preview what you plan to save. Clear, varied singing matters more than extending the material simply to reach the maximum length.
Cloning a voice for singing
This workflow trains voices for singing. Use a trained voice with song covers or original songs; it is not a text-to-speech or narration tool.
Choose 2–30 minutes of clear solo singing, with 10–20 minutes recommended. Quality and variety matter: avoid padding the material with long silence, talking or repeated noisy passages.
Yes. Your singing voices belong to your account library. Once a voice is ready, you can select it in another cover or original-song workflow without training it again.
Yes. Name each voice and add an avatar in the library so it is easy to recognize when choosing a singer.
No. Saving a recording does not automatically start paid training. Training is a separate purchase, priced by the duration of your recordings in minutes, rounded up. Recordings with Extract vocals turned on add a per-minute extraction charge. Song generation and voice conversion have their own charges. Review the relevant quote before submitting each task.
Use your own voice or a voice you have explicit permission to train and use. You must confirm the required permissions when adding the recordings.
Yes. Add the recording and turn on Extract vocals for it. The voice is isolated from the music before training. Extraction adds a per-minute charge, shown in the training quote, and works on recordings up to 10 minutes long. Dry solo singing still gives the cleanest result, so leave the option off for those recordings.
Training runs asynchronously, and longer recordings can take over an hour. There is no guaranteed completion time. You can leave this page and return to your library to check progress; use the voice when its training status is ready.
Results depend on your training recordings and the source vocal performance. Conversion follows the source singer's notes, timing and phrasing while changing the vocal timbre. Consider your natural range, listen to the result, and adjust the song or conversion pitch when needed.
Your singing voices and avatars stay in your private account library. Manage their names, pictures and training status there, or remove a voice you no longer need. A voice that is currently training or being used in a conversion can be removed after its running tasks finish.
Put your singing voice into a song
Build the singing voice for your next song
Start with clear solo singing, train your private voice, and return to it whenever you create your next cover or original song.