It separates the vocal from the instrumental, writes the lyrics with the timing of every syllable and measures the pitch of every note. This takes a while — upload the file, go do something else, and when you come back the karaoke is already in your library, ready to sing.
3 songs a month on the free plan. No card to try it.
is how long processing takes today, measured, against the length of the song. No promised deadline — but nothing hidden from you either.
upload audio (you get an animated stage with mic and crowd) or video (the lyrics go over your clip).
the final score, with a nickname. From "A joke" to "Legend".
01 what happens to the file
A separation model pulls the vocal out of the instrumental. What is left becomes the backing track you sing over; the isolated vocal becomes the raw material for the next two steps.
First it looks up the officially published lyrics and checks whether they match your audio, showing the match percentage. If they match, it uses them; if not, it says so and uses the transcription anyway — always with the timing of every syllable, which is what makes the word light up at the right moment instead of the whole line appearing at once.
Pitch comes from the isolated vocal signal, syllable by syllable. A spoken or shouted passage does not become a pitched note — it becomes a rhythm note, which you hit by landing on the beat.
Mic on, notes running across the screen, score climbing with your pitch. At the end you get a 0-to-10 score and a vertical video ready to post — no video editor involved.
02 the final score
Every syllable earns points for pitch and for landing on time. It adds up to a 0-to-10 score — with a nickname, which is the part the group remembers.
We simulated someone who does not know the song: they sing, they hold notes, they come in off the beat. Against five real songs, on the default setting, this is what they score:
Someone singing it right scores 100% in all three. Rhythm only forgives pitch on purpose — it is the mode for singing along without being judged.
03 plans
to see if it is any good
or R$ 149 a year — two months free
04 before you subscribe
There is no song search here, and there never will be. StarSinger distributes nothing: you bring a file that is already yours and we process it.
The result stays in your private library, visible only to you. That is the model of an editing tool, not a streaming service.
05 frequently asked questions
We do not promise a fixed time because it would be a lie — it varies with file size and server queue. As a yardstick: today we measure about 20 times the length of the song itself. Upload it and go do something else: you do not have to keep the tab open waiting, the result will be waiting in your library when you come back.
You can. If you upload audio, the karaoke gets an animated stage with a mic and a crowd. If you upload video, the lyrics go over your own clip instead of the stage. It accepts mp3, wav, m4a, ogg and flac for audio, or mp4, mov, webm and mkv for video — up to 400 MB per file, one song processing at a time.
No. There is no public catalogue on StarSinger and no search across other people's songs. What you upload stays only in your private library — that is the model of an editing tool, not a streaming service.
For lyrics, the system looks up the officially published version and checks whether it matches the audio, showing the match percentage. If it does not match, it says so and uses the automatic transcription instead. Pitch is measured from the isolated vocal signal, so on songs with a lot of instrumentation over the voice it errs more. Good enough to sing with friends — not built for competition.
You can. The subscription is month to month (or yearly, if you pick that plan) with no lock-in. If you cancel, you keep your library and the Free plan rules apply from the next cycle.
Upload the file, go do something else. When you come back, the mic can take the stage.