What your dubbing score measures
The score compares the shape and length of your delivery with the reference. It is a practice aid, not a verdict on your voice or acting ability.
Record a line, replay your take, then choose one thing to adjust. If you have not played yet, the first-session guide explains the controls.
Three measurements, three different questions
Rhythm: where does your delivery rise and fall?
The game compares changes in audio energy across the line. Stressed words, pauses and a drawn-out ending change that shape. Each recording's volume is normalized, so simply speaking louder is not a reliable way to improve this score.
Listen for one distinctive moment in the reference: a pause before the last word, a quick opening, or an emphasized middle syllable. Try matching that moment in your next take while keeping your own voice.
Duration: how long did the spoken part last?
The game trims leading and trailing silence, then compares the remaining duration with the reference. A rushed or unfinished line can lose duration points. Silence before speaking is not the same measurement as the length of the spoken line, although coming in late can still spoil the finished scene.
If this score is low, check whether you finished the final word or stretched a phrase too far. Replaying your recording is more useful than guessing from the overall score.
Coverage: is your take broken up by pauses?
Coverage estimates how much of your trimmed take contains voiced audio. It does not recognize the script or count how many words you completed. A short, cleanly spoken fragment can have high coverage and low duration at the same time.
If coverage is low, listen for unintended gaps inside a phrase. Do not remove every pause: some pauses are part of the reference performance. Compare the sound as well as the number.
How the overall number is calculated
The current scoring version weights rhythm at 58%, duration at 32% and coverage at 10%. It rounds the result and limits a scored take to 10–99. A 100 is not available in this version.
For illustration, component values of 80% rhythm, 90% duration and 100% coverage give 85 after weighting and rounding. This is an arithmetic example, not a recorded player result.
Silence, very quiet input, a brief bump or steady noise may trigger the no-voice check instead of a displayed score. If the game cannot hear a real take, fix the microphone input before trying to optimize your performance.
Use one retry to answer one question
- Replay the reference. Pick a pause or stressed word that you can hear clearly.
- Record the complete line. Keep a comfortable volume and finish the ending.
- Listen to Your take. Check for missing words, unwanted room audio and timing that feels wrong.
- Change one thing. Adjust that pause, phrase length or emphasis, then compare the next take.
A higher score is useful feedback when the recording also sounds better to you. If a take you prefer scores slightly lower, you can still keep it for the finished dub.
What the score does not tell you
- It does not recognize your words, judge pronunciation or test language fluency.
- It does not compare pitch, accent, vocal identity or how convincingly you imitate an actor.
- It does not measure whether the emotion or joke lands. Listen to the whole scene for that.
- Background sound, microphone processing and noisy reference audio can affect the comparison.
The comparison runs in your browser; it does not need to send your microphone recording to a speech-recognition service. Optional challenge uploads and site analytics are explained separately in Privacy.