Your own voice is created in "Voices" from a single audio recording. The cleaner the recording, the closer the voice: a clone copies not only the timbre but also noise, echo and the manner of speech from the sample.
Recording requirements #
- One person, no music, no second voice, no noise. A quiet room without echo.
- From 30 seconds to 10 minutes. The first 60 seconds of clean speech are used for the clone, so the most important part is the beginning.
- MP3 or WAV. File size within your plan's limit.
- Speak naturally, the way you want to hear it in the translation. Pauses, intonation and energy carry over into the clone.
- Set the recording language: the one the person speaks in the sample.
There is no in-browser microphone recording. Record on your phone or computer and upload the file. Video files are not accepted: extract the audio with any converter.
What happens after upload #
- The voice gets the "Processing" status: we separate speech from the background and level the volume. Usually it takes minutes.
- "Active": the voice can be picked under "My voices" when uploading a video and in the editor.
- The clone itself is created at the voice engine on the first dubbing or the first listen.
You can hear how the voice sounds: the ▶ button on its card in "Voices" (a phrase in the recording language) and ▶ next to the voice in the editor (a phrase from your translation, including the "no accent" option). The first time the sample takes up to 15 seconds.
"Error" means speech could not be separated. Check that the file contains a voice and is at least 30 seconds long, then upload again. If "Processing" lasts more than an hour, write to support.
Limits #
Own voices are available on Pro (up to 5) and Business (up to 10). On Free and Start the "Voices" page shows only clones from your videos. A voice can be deleted from the menu on its card; in older translations it is replaced with a standard voice.
About the "Remove accent" switch see Accent in the translation.