Redub a video into another language

Pro feature · Stock neural voices · No cloning required · Pricing · Voice Studio · How we test

Redub video audio online with VoxCraft: upload a clip, choose a target language and neural voice, and download a new dubbed MP4 that keeps the original picture. The pipeline extracts audio, transcribes speech, translates the script, lets you review and correct every line, re-voices it, then adds the new track back onto the video, with the original background music kept underneath. No voice cloning or local GPU needed.

Built for creators who need an Urdu, Hindi, Tamil, Telugu, Arabic, or English version of an existing explainer without re-recording. Voices come from the same library as Voice Studio. If the source is noisy, tick Denoise audio before transcribing on the form, or clean the track first with Remove background noise.

Redub is not automatic lip-sync acting and not a clone of the original speaker in Phase 1. It replaces the spoken track with a clear stock neural voice in the target language. That is the right tool when localization speed matters more than matching the original timbre. For same-language retakes in your voice, see Voice cloning (coming soon).

Character usage follows the same TTS quota rules as Studio. Long videos consume more quota, so prefer clear speech, moderate length, and a short test clip when you are trying a new language pair. See what counts toward your limits below.

How it works

1. Upload an MP4, MOV, WebM or MKV (up to about 50 MB / ~10 minutes).
2. Set your options: source and target language, voice, speech rate, background music and transcription engine (all explained below).
3. Tap Redub video. VoxCraft extracts the audio, transcribes it with timestamps, and translates each line.
4. Review the lines. Press Play to hear each one, fix wrong words, adjust the timing, merge or skip lines.
5. Tap Generate dubbed video. The new voice is created, placed on the original timeline and mixed over the background music.
6. Download the dubbed MP4 or the audio-only track. Not happy? Change the voice, speed or music and generate again without uploading.

Prefer one tap? Untick Review and edit the transcript and VoxCraft does everything in a single step.

Options before you start

• Source language: the language spoken in the video. Choose it explicitly for the best result (for example Hindi for Hindi/Urdu clips); leave Auto only when speech is very clear.
• Transcription engine: Auto is recommended. For Hindi and Urdu it uses Whisper to find when each line is spoken and Google Speech to write what was said; for other languages it uses Whisper first and falls back to Google. You can also force Whisper only or Google only.
• Target language and voice: any of the 130+ stock voices. Pick a voice that matches the target language.
• Second voice (optional): alternates voices line by line, useful for two-speaker clips.
• Speech rate: 50% to 150% of the voice’s normal speed.
• Match original video length: keeps each line on the original timeline at a natural pace. A line may borrow the pause after it and is only lightly sped up if it still runs long.
• Background music: Keep (normal volume), Keep (quiet) or Remove. See below.
• Denoise audio before transcribing: cleans the extracted audio so noisy speech is recognised better. Only the copy used for transcription is cleaned; your video and its picture are not changed. It makes the first step slower.
• Review and edit the transcript: on by default and recommended.

Review and fix every line

No speech recognition is perfect on noisy audio, mixed languages, numbers or product names. The review screen shows each spoken line as a card so you can correct it before any voice is made:

• ▶ Play plays that exact stretch of your original video so you can hear what was said.
• What was heard: edit the text if the words are wrong.
• Re-translate from edited text: after fixing the heard text, refresh the translation for that line.
• Voiced line: the exact words the new voice will say. Edit them freely.
• Start / End (seconds): move a line earlier or later if it is out of sync with the speaker.
• Merge with next: joins two lines when one sentence was split in the wrong place.
• Skip line: leaves that stretch silent, for example for music, noise or a line you do not want voiced.

Keeping the background music

With Keep selected, VoxCraft uses AI speech isolation to separate the original voice from the music and ambient sound, then mixes that background back in under the new voice. The music is turned down while the new voice speaks and comes back up in the pauses.

• Keep (normal volume) is the default. Keep (quiet) hides any leftover traces of the original voice better. Remove gives the new voice only.
• This is a separation trick, not a studio stem: faint traces of the original voice can remain, especially in echoey rooms.
• It works on clips up to about 6 minutes. If there is no music in the clip, or separation is not possible, the dub is made with the new voice only and the result note tells you why.

Controls after your video is dubbed

• Preview the dubbed video in the page, then Download dubbed MP4 or Download audio only (the new voice with the music, ready to re-mux in your own editor).
• Show transcript & translation lists the original and voiced text. The note under the video says which engine was used, how many lines were timed and what happened with the music.
• Try another voice, speed or music level: change them in the form, edit any line and tap Generate dubbed video again. Your review session stays available for 45 minutes, so there is no need to upload again.
• Changing the target language needs a fresh transcription: press Redub video again so the lines are translated into the new language.
• After 45 minutes the review session, including your uploaded video, is deleted automatically. Upload the video again if you need more changes.

What counts toward your limits

• Each Redub video run counts once toward your daily and monthly redub allowance, when the transcription finishes. Editing lines and generating again does not add another run.
• TTS characters (your voiced text) are counted every time you tap Generate, the same way as in Studio. Skipped lines use no characters.
• If a redub would go over your monthly TTS quota, you are told before any voice is made.

Who uses video redub

Tips for better redubs

• Always use the review step on noisy or Hinglish clips. Play each line, then fix numbers, names and product terms by hand.
• Clear speech transcribes better than music-heavy or noisy tracks. Tick Denoise audio before transcribing if the room tone is strong.
• If a voiced line starts too early or late, adjust its Start/End seconds. If one sentence was cut in two, use Merge with next.
• Music-heavy clip? Try Keep (quiet) so any leftover voice traces stay hidden.
• If source and target are the same language, translation is skipped and only the voice is replaced.
• Match the target voice’s language to the target language for natural pronunciation.
• Start with a 30 to 60 second sample before redubbing a full 10-minute lesson.
• For script-first narration (no existing video), use Voice Studio instead of redub.

Frequently asked questions

Related VoxCraft tools