About Speaker match

Speaker match applies your voice characteristics (tone, accent patterns, speaking style) to the translated audio, so listeners hear the translation in a voice resembling yours rather than a generic preset voice.

How it works

DeepL uses a rolling sample of your speech from the current meeting as the reference for synthesis. The sample is held in memory only and is discarded when the session ends.

• No recording. There is no record step, and no audio is recorded.
• No upload. There is no upload step. You do not submit a voice sample.
• No storage. No audio is stored after the session.
• No reuse across sessions. Nothing from one meeting is carried into another.
• No training. DeepL does not use your voice, your meeting audio, transcripts, or translations to train or fine-tune translation or synthesis models. No sub-processor is instructed to do so either.
• No voice profile. Speaker match does not build a voice profile that is kept after your meeting ends.
• No identification. Speaker match matches how you sound. It does not identify who you are.

Available platforms

DeepL Voice for Online Meetings

• DeepL Voice Desktop App
  To enable or disable Speaker match in the DeepL Voice Desktop App, go to: Settings > General > What others hear.

DeepL Voice for In-Person Conversations

• DeepL Web App
  Go to: Toolbar > Audio > Your voice and select Speaker match
• DeepL mobile app
  Tap > Your voice and select Speaker match
   

Available languages

Speaker match is supported exclusively on native DeepL text-to-speech (TTS) models. If a language relies on a third-party provider, Speaker match is unavailable and a standard fallback preset will be used instead.

See the full list of available languages - DeepL Voice languages.