Original-voice dubbing and voice review

Check the words and speakers first

A natural-sounding voice can still say the wrong words. Review the transcript, translation, names and speaker labels before generating the final speech track. Use source material and reference voices you have permission to use.

For a conversation, keep each speaker’s text and voice assignment distinct. A single voice reference should not be treated as the correct voice for every person in the recording.

Generate and inspect a voice version

1. Open the translation or voice task for the selected video or clip. Confirm the target language and the intended speaker.
2. Review the reference and speaker assignment available in that task. Avoid reference passages dominated by overlap, music or another speaker.
3. Generate a short representative passage and listen for pronunciation, voice identity, pace and completeness.
4. Correct the text or assignment when needed, generate the updated speech, and listen again before running lip sync.

What to check when a voice sounds wrong

• Wrong person: inspect speaker labels and the voice reference rather than trying a different face crop.
• A name is pronounced incorrectly: check the spelling and the pronunciation or terminology controls available in the task.
• Speech sounds rushed or clipped: compare the translated wording and pacing with the source passage.
• Overlapping speakers: review the source and assignments individually; voice generation alone does not resolve every overlap.

Review audio before lip sync

Lip sync follows the audio you provide; it does not repair mistranslated dialogue or an incorrect voice assignment. Confirm the speech track before spending time on mouth movement and final visual review.

Related guides

• Translate a video from import to export
• Multi-speaker and multi-angle lip sync
• Task history, queues and troubleshooting