AI Voice Cloning: Keep a Character's Voice, or Use Your Own

Quick answer: SceneMixer clones a voice in two places. When you change a line in a finished clip, Redub’s Original voice clones each speaker’s voice from that clip and uses the clone for that redub only, so the new words sound like the old ones. And on a paid plan you can upload a recording for a character, which the video model then uses as that character’s voice in every segment they speak in. Both are for voices you have the right to use; every other voice comes from the voice library or a written description.
Still from “Sky Rider's War”, an AI short drama generated end-to-end by SceneMixer (10 characters, 1 episode)
SampleSky Rider's WarShot on Film1 episodeProduced end-to-end by SceneMixer from the scriptWatch the episodes

Try it on a clip you already have: Edit this clip → Redub, change one word, leave the pill on Original voice, choose Lip sync and press Redub into a new version.

What does Original voice clone?

Original voice is a choice in Redub, the panel for changing what is said in a finished clip. SceneMixer takes the clip’s own soundtrack — from the version the video model generated, not from a later redub — finds the stretches where each person speaks, and builds a temporary voice from them. Your new lines are spoken in that voice and the mouths are moved to fit. The temporary voice is used for this redub only and deleted afterwards; the next redub clones again.

In the panel, each line’s voice pill starts as Original voice when the clip allows it. Opening the pill shows This clip's original voice as the first row, then the voice library.

  1. Select the segment and open Edit this clip → Redub.
  2. Change the words you want; leave the others as they are.
  3. Leave the voice pill on Original voice.
  4. Choose Lip sync.
  5. Press Redub into a new version. A new version is saved, and the clip you started from stays in the version list.

Change a line after the video is made · Lip sync or Re-perform

Which clips and languages can use it?

  • Seven dialogue languages: Chinese (dialects such as Cantonese included), English, French, German, Japanese, Korean and Russian. In any other language the pill offers library voices only.
  • Several speakers: each person gets a clone of their own, cut from the parts where they speak.
  • Someone who says very little: if a character speaks for less than about three seconds in this clip, SceneMixer adds their speech from other clips of the same project. If there is still too little, the redub fails and the credits come back.
  • The clip has to lead back to its generated original. Trims, AI visuals, earlier redubs and enhancement keep that link. When a clip has none, Original voice is not offered.
  • Lip sync only. Re-perform acts the clip again with a library voice.

How much does voice cloning cost?

The same as any Lip sync redub: 2 credits per second of the whole clip, shown on the button before you press it. Cloning adds nothing to that price. If the clone cannot be made — too little speech, or the provider’s check turns it down — the redub fails, the credits come back, and you can pick a library voice and try again. What each clip fix costs

Can I clone my own voice for a character?

Yes, on a paid plan. In step 2, open the character and use Upload & Replace in its voice section to upload a recording; the voice then reads Uploaded audio. From then on the video model takes that recording as the character’s voice reference in every segment they speak in, so new segments sound like your sample. Uploading costs no credits. Upload only your own voice, or a voice you have the right to use. How character voices work

What about a remake?

In a remake, Swap keeps the sound of the clip you uploaded, because it is your clip itself, not a copy. If you change a line in Swap, the part with the changed line is generated again and AI re-voices everyone in that part — a new performance rather than a clone. Remake a video with AI

What SceneMixer does not do

  • No separate voice cloner. You cannot upload a voice and download speech from it; cloning happens inside a redub or a generation.
  • No saved clones. The Original voice clone is not kept as a library voice.
  • No voices you have no right to use. A public figure, a client or a friend without their permission is not allowed, whatever the tool.
WayWhereCostLasts for
Original voiceEdit this clip → Redub, with Lip syncThe Lip sync price, 2 credits a secondThat redub
Your own recordingStep 2, the character’s voice, Upload & Replace (paid plan)No creditsEvery segment the character speaks in
Voice libraryThe voice pill in Redub, or the character’s voiceFree to pickThat redub, or the character
Written descriptionThe character’s voice descriptionFreeEvery segment with no audio attached

FAQ

Does SceneMixer have voice cloning?

Yes. SceneMixer clones a voice in Redub (Original voice, taken from the clip being redubbed), and paid plans can upload a recording as a character's voice. Both are for voices you have the right to use.

Is the cloned voice saved for later?

No. It is built for one redub and deleted afterwards; the next redub clones it again from that clip.

Why don't I see Original voice?

The project's dialogue language is not one of the seven it supports, the clip cannot be traced back to its generated original, or you picked Re-perform. The pill then offers library voices.

Does it work when two or three people talk?

Yes. Each speaker is cloned from their own parts of the clip. If one of them says too little, even after their lines from other clips of the project are added, the redub fails and the credits come back.

Can I clone a celebrity's voice?

No. Upload only your own voice or one you have permission to use.

Can I type text and get audio in a cloned voice?

No. There is no separate text-to-speech download; cloned voices are used inside a redub of your clip.

Does the new line keep the background sound?

No. Lip sync replaces the clip's soundtrack with the new voices, and the episode's music comes back under it on its own. Re-perform keeps a full take but uses a library voice.

Keep the voice, change the words

Open a finished clip, change one line and leave the pill on Original voice.

Open SceneMixer

Updated November 15, 2026