Dub a Scene Without Losing the Sync

The problem

Dubbing used to mean a studio, an actor and a week. Now it is a button. The button is also where most dubs go wrong: the translation runs long, the voice speeds up to fit, and the mouth stops matching. Good dubbing is a timing problem before it is a voice problem.

One thing to know before you start: ElevenLabs Dubbing translates and re-voices your audio. It does not move the lips. Lip sync is a separate step with a separate tool, and you only need it for shots where the mouth is clearly visible.

Two dubbing modes

• Dubbing v2 (default). Automatic. Upload, pick languages, download. No in-app editing. Fast and often good enough for wide shots, voiceover and anything where the mouth is not the focus.
• Dubbing Studio (choose "Dubbing v1" then "Dubbing project"). Gives you a timeline with speaker tracks, editable transcripts and translations, per-clip regeneration and per-clip voice settings. Use it whenever you need to control timing. It is in maintenance mode, so do not expect new features, but it works.

Steps

1. Cut the scene first, dub second. Never dub a rough cut. Every re-edit breaks the timing you fixed.
2. Upload with separated audio if you can. If you have dialogue and music or effects on separate tracks, upload the dialogue only and add the rest back in the edit. Cleaner input, cleaner voice.
3. Fix the translation for length, not accuracy. Open the speaker cards in Dubbing Studio and read the translation out loud against the original. German runs about 20 to 30 percent longer than English. Cut words until the line fits the same breath. A shorter faithful line beats a complete rushed one.
4. Understand Fixed vs Dynamic generation. By default, regenerated clips are Fixed: the clip keeps its duration and the speech speeds up or slows down to fit. That is the chipmunk effect. Right click a clip and switch to Dynamic when you want natural pace and can afford the clip to grow, but check it does not collide with the next line.
5. Use a clip clone for the best take. If one clip nails the voice, right click it, "Create Voice from Selection", and assign that voice to the whole track. Consistency across the scene improves immediately.
6. Lip sync only the close-ups. Export the dubbed audio, then run only the shots where lips are clearly visible through a lip-sync model (Runway Act-Two drives a performance from a video, other tools take video plus audio). Wide shots and over-the-shoulders do not need it and will look worse if you force it.

Worked example

A 40-second two-hander in English, target German.

• Cut locked. Dialogue exported as its own track.
• Dubbing Studio project created. Speaker A and B detected.
• Line 4 translation runs long: "Ich habe dir doch gesagt, dass ich hier sein würde" becomes "Ich hab's dir gesagt. Ich bin hier." Same meaning, same breath.
• Line 7 sounds rushed on Fixed. Switched to Dynamic, clip grows by 0.4 seconds, still clear of line 8.
• Clip clone from line 2 assigned to Speaker A's whole track.
• Export WAV per speaker. Two close-ups run through lip sync. Everything back into the edit, music and effects underneath.