Start with the song: paste a link or upload an MP3 and press Process. The karaoke studio opens with the voice and the band apart, and “Add a track in your voice” re-sings the lead from 10 to 30 seconds of you. Then mix, compare and download your cover.
Pro allows 15 min on Standard · 6 min on High Quality
An AI song cover keeps everything about a song except the singer's voice. Mazmazika makes one in two steps. First the vocal remover separates the lead vocal from the band. Then a singing voice conversion model re-sings that vocal with the timbre of your voice, note for note and word for word, and places it on a new track beside the original.
You only need a short sample of your own voice, singing or speaking. The cover is made inside the karaoke studio, so you can compare it with the original singer, mix it with the band, add your own recorded takes and download the result.
20 seconds of a royalty-free song, split by the Ultra engine on this page, then its lead vocal re-sung by AI voice cloning from a 20-second voice sample, once with each engine. Switch tracks while it plays: the words and the melody stay the singer’s. Every track plays at the same loudness.
Song: “Eternal Light” by mahbubjoarder (Pixabay Content License), separated on this site; only the loudness is matched. The cloned voice is a Mazmazika team member’s, used with their permission.
Read live from the current plan settings, so this section always matches what the tool enforces today.
See the plans and current prices
Voice conversion changes who is singing, not what is sung. The melody, the rhythm, the phrasing and the words all come from the original vocal; only the timbre becomes yours. That is why the cover works for songs in any language, and why it can sing notes and runs you have never practised.
Each conversion becomes a new track, up to eight voice tracks per song (clones and recorded takes together), so you can try both engines, several octaves or different numbers of steps and keep the best. More steps give more detail and take proportionally longer.
The sample is stored in your account only after you confirm that it is your own voice or that you have the speaker's permission. It is used only to make your voice tracks, and you can replace or delete it at any time. The copy sent to the GPU for a job is deleted with that job.
Only with that person's permission. The tool is built for your own voice: you confirm that the sample is yours, or that the speaker agreed, before it is stored.
Any. The conversion changes only the voice, so the words stay exactly as the original singer sang them. If the words of one song come out unclear, try the other engine: the two learned from different data.
The split takes under a minute for a typical four-minute song on Ultra, and each voice track about half a minute to a minute more. SoulX-Singer takes about twice as long as Seed-VC.
10 to 30 seconds of your voice alone, singing or speaking, recorded in a quiet room without music or effects.
The song and its recording keep their copyright, so publishing or streaming a cover needs the rights holder's permission. Listening, practice and sharing privately with friends are the intended use.
