To retain a speaker’s recognizable character in another language, use voice-matched dubbing: transcribe the source, translate by segment, synthesize each speaker from reference audio, and align the new speech to the original timing. The result is generated speech, not the untouched original recording. Review it carefully and use a person’s voice only with permission.

What “keep the original voice” really means

The original waveform cannot simply speak a different language. Voice-matched dubbing instead uses clean portions of the source performance as references for synthetic speech. The goal is continuity in speaker identity and delivery, not a claim that the new audio is the original recording.

If preserving the original video frames matters more than changing mouth movement, first compare audio dubbing and lip-sync dubbing.

1. Start with clean, authorized source audio

Use a file you own or are allowed to process, and obtain the necessary consent to synthesize each voice. Clear turns, limited background music, and little overlap give transcription and speaker detection better material to work with.

TransCast accepts MP4, MOV, WebM, and MKV files up to 500 MB and one hour. It does not import video URLs.

2. Choose one target language

Each job produces one translated version. Decide on the audience before processing so the reviewer can judge terminology, register, names, numbers, and units for that market.

3. Lock terms and pronunciations

Add names, brands, acronyms, and product language to the glossary. Each entry separates the source term, the exact subtitle display, and the text passed to speech synthesis for pronunciation.

This is particularly important for lessons and recurring material. See the online course localization workflow for a repeatable approach.

4. Generate speaker-aware speech

TransCast transcribes speech into timed, speaker-labeled segments, extracts reference audio, translates each segment, and generates new speech from the relevant reference. It then adjusts wording and audio timing to fit the original segment windows.

For conversations, review the extra risks in multi-speaker interview dubbing. There is currently no manual speaker-assignment editor.

5. Review the video and bilingual SRT

Listen for meaning, missing words, pronunciation, speaker changes, unnatural speed, and lines that land after the related action. Then read both lines of the bilingual SRT against the audio. You can keep the SRT separate or burn both languages into the delivered MP4.

  • You have rights to the source video and permission to synthesize the voices.
  • A fluent reviewer has checked the target language.
  • Names and technical terms match the intended spelling and pronunciation.
  • Every speaker change and fast exchange has been inspected.
  • The audience can tell when disclosure of synthetic media is appropriate.

Create a voice-matched dub.