How Evocal dubs a video
What happens between the upload and the finished dub, and the rules the engine keeps at every step.
The stages of a dub
The project page shows which stage a dub is in and how far it went. In order:
| Stage | What happens |
|---|---|
| In line | The dub waits for its turn. Each plan runs a set number of dubs at the same time; the next one starts when one ends. |
| Reading the file | The video or audio is read: its length, its sound tracks and its language. |
| Separating voice and music | The speech is taken apart from the music and the other sounds. |
| Transcribing | The speech is written down, with the time each word was said. |
| Translating | The text is translated thought by thought, sized to the time the speaker took. |
| Making the voices | The new speech is made in each speaker's own voice. |
| Mixing | The new speech is placed on the video's clock and mixed over the music and background, when you kept them. |
| Packing the files | The video, the audio files and the subtitles are prepared for download. |
The rules the engine keeps
The voice comes from your file
There is no voice to record or train first. Evocal takes each voice from the speech in the file you upload, and each person speaking is dubbed in their own voice.
The video's own clock
Each thought is placed where it was said in the original. A silence in the original stays a silence in the dub, a hesitation stays a hesitation, and the voice reaches a click when the cursor does. The dub does not speed up to fit or fill pauses that were not there.
Whole thoughts, never patched words
Speech is made one whole thought at a time. When you change a sentence in the editor, the whole segment is made again in the same voice; a single word is never cut in.
It checks its own work
Before a dub reaches you, the engine checks the result against the text and the original, and marks the segments worth a listen. Those marks are the alerts in the editor.
Every dub is marked as AI
The exported files carry AI marking: provenance information in the file and a watermark in the audio, neither of which you hear. The terms ask you to tell your audience that the dub was made with AI. Dubs made with the free trial also carry a visible watermark.
What it does not do
- Lip sync: the new speech follows the timing of the speech, not the mouth.
- Dub from a link: you upload the file itself, so the project starts from a video you hold.
- Keep a library of voices: each dub takes its voices from its own file.