Lore

11 Labs voice changer on the extracted audio track

# 11 Labs voice changer on the extracted audio track

AI avatars often sound "a little bit tiny or a little bit AI". This technique replaces the avatar's synthetic voice with a converted voice. The converted voice keeps the intonation and pauses of the source audio, so the delivery stays human. It is one of the two audio fixes in Audio realism as the last step that makes AI UGC pass.

Procedure

  1. Generate the talking video, see Turning image variations into talking UGC video ads.
  2. Extract the audio track from the AI video. Premiere Pro, CapCut or a free online extractor all work, and ChatGPT can walk you through it.
  3. Upload the audio to 11 Labs and switch it to the voice you want, using the voice changer (speech-to-speech) feature.
  4. Put the converted audio back under the video.

Variant: your own performance

The source audio does not have to come from the AI video. In the demo, the presenter records a line in their own voice and converts it to a female voice. The cadence of the human take carries over to the new voice. Playing the original and converted tracks at the same time shows that the intonation and pauses match.

When to use it