Niche-Knack

Desktop · 0:53 video

Preview a change to the spoken words

Edit a short line, hear a local voice preview, then explicitly accept or undo the audio replacement.

Before you start: A transcript, the optional Pocket TTS pack, and permission to use the speaker’s voice.

Download videoDownload captions

Step by step

  1. Select the words to replace

    Select a short passage in Transcript. Open Speakers and wording → Change spoken words…. This is separate from Correct word, which edits transcript text only.

  2. Set the line and reference

    Enter New spoken wording. Listen to a clear 2–10 second reference from the same speaker, and confirm that you have permission to use their voice.

  3. Generate and listen

    Choose Generate voice preview, then Listen to voice preview. The preview preserves the selected passage’s duration and includes the edit fades. Generate again after changing the wording.

  4. Accept only a useful result

    Accept audio replacement keeps the generated audio and matching words as an undoable edit. Ctrl + Z restores the original audio and transcript together.

Clip Dr. showing listen to the generated voice preview

Listen to the generated voice preview. Open full size

  1. The revised spoken line is separate from the original wording.
  2. Listen to a clear reference with permission.
  3. Listen to the generated preview.
  4. Accept only after reviewing the result.
Read the video instructions
  1. This example uses a synthetic voice. Use a real speaker’s voice only with their permission.
  2. 1. Select the spoken passage. Open Speakers and wording → Change spoken words…
  3. 2. Edit the line. Check the voice reference and confirm permission to use that voice.
  4. 3. Generate voice preview. Generation runs locally on the CPU.
  5. Processing runs locally. Wait for the result before continuing.
  6. The result is ready to review.
  7. Listen to the preview before accepting. The passage keeps its length and edit fades.
  8. 4. Accept audio replacement keeps the preview. Ctrl + Z restores the original audio and words.
  9. Keep only a preview you are happy with. Correct word remains the simpler text-only correction tool.
Audio credits

LibriSpeech excerpts 1688-142285-0007 and 1272-128104-0000, collected by Vassil Panayotov, Guoguo Chen, Daniel Povey, and Sanjeev Khudanpur from LibriVox, are used under CC BY 4.0. Excerpts have been trimmed, resampled, padded, or processed. The voice-replacement example uses an original synthetic voice.