Dubbing a finished video with ElevenLabs

TL;DR: Select a finished video clip in Premiere, choose the ElevenLabs Dubbing model in modelBridge, type the target language code, generate, preview, and import. The dubbed version comes back as a complete video file — original cut, replaced audio — that lands at the exact same timecode with one click.

Why this workflow exists

Voice-over replaces narration. Dubbing localizes a finished piece of content for another language and audience. That means the job is not just voice generation — it includes transcription, translation, and timing against the original performance.

Use this workflow when you have a locked or near-locked edit and need a version for another market, language, or audience. The modelBridge integration makes the difference: what would normally be a 12-step export → upload → wait → download → re-import → re-align cycle becomes a single replace on your existing timeline cut.

What the model actually does

ElevenLabs Dubbing on fal.ai receives your video or audio file, transcribes the speech, translates it, and returns a dubbed video file with the audio track replaced. The output is always a video file — even if you send audio only, you get video back. This means the dubbed result goes directly on a video track in your timeline, not an audio track.

What makes a good dubbing source

Good candidates:

Clean dialogue with limited background music under speech
Single speaker or clearly separated speakers
Slower pacing with natural pauses
Wide or medium shots where lip-sync mismatch is less visible
Shorter segments — better control, easier to regenerate problem sections

Harder to dub well:

Dense background music competing with dialogue
Fast, overlapping speech
Close-up talking heads with visible mouth movement throughout
Legal copy, brand terms, humor, or slogans that require cultural adaptation

Step 1 — Select your source clip

Click your finished video clip on the Premiere timeline. modelBridge detects it automatically and populates the video field. If you select an audio-only clip, it populates the audio field instead — you still get a video file back.

Step 2 — Find the Dubbing model

In modelBridge Browse, click the Dubbing filter chip or search “dubbing.” Select the ElevenLabs Dubbing model.

Dubbing is a separate filter from Voice-over — it is not listed there. Voice-over creates or replaces a narration track from text. Dubbing localizes an existing performance into another language. Different job, different model, different filter.

Step 3 — Configure

Target language (required) — enter an ISO 639-1 language code. Common codes: es (Spanish), fr (French), de (German), ja (Japanese), pt (Portuguese), zh (Chinese). The field accepts free text — type the code exactly. Typos will fail at the API level with no warning, so double-check the code before generating.

Source language (optional) — auto-detected if not set. Set it explicitly if auto-detection gives wrong results.

Number of speakers (optional) — auto-detected. Set manually if the model misidentifies speaker count in complex scenes.

Highest resolution — on by default. Keep it on.

Step 4 — Generate and preview

Click Generate. modelBridge uploads your source clip to fal.ai. Generation time depends on clip length.

The result appears in the modelBridge preview panel as a video with playback controls — not an audio player. Click the preview card to open it in Premiere’s Source Monitor for full-screen playback.

Before importing, review:

Is the translation accurate? Check product names, brand terms, legal copy, and slogans manually
Does the pacing feel natural in the target language? Some languages run longer or shorter than the source
Does the dubbed voice work against visible mouth movement? Check this most carefully on close-up talking heads
Are speakers distinguishable if the source had multiple voices?

Step 5 — Import

Click Import to Timeline.

Replace on Timeline: If your source clip came from the timeline, the dubbed video replaces it at the exact same timecode, same position, same duration. The original clip — including its audio track — is overwritten. There is no option to keep the original video and swap only the audio, since the dubbed result is a complete video file.

Insert at playhead: Also available if you want to place the dubbed version on a separate track without removing the original.

Workflows unique to modelBridge

In-place timecode dubbing Select a clip at 00:01:23:00 → dub to Spanish → Import to Timeline. The dubbed video lands at exactly 00:01:23:00 with the same duration. No manual repositioning. The cut stays intact. This replaces what would otherwise be: export → upload → wait → download → import → drag to timecode → scale → check sync.

A/B language comparison on timeline Select a dialogue clip → dub to Spanish → Insert (not replace) → the dubbed version lands at the playhead on a separate track. Toggle track visibility to compare sync, lip movement, and pacing between English and Spanish. Repeat for French and German. Build a multi-language comparison without leaving Premiere.

Isolation → dub pipeline Source clip has dialogue over music and SFX → run through Audio Isolation in modelBridge to get a clean dialogue track → dub the isolated audio → re-layer with the original music/SFX in Premiere. Both models run in the same panel without any export cycles between them.

Per-clip dubbing for interview edits An interview cut with 12 clips from different speakers. Select each clip individually → dub → replace. Each clip gets dubbed in context, preserving the edit timing. Without modelBridge: 12 separate export-upload-download-import- realign cycles.

Sequence versioning for localization deliverables Duplicate the sequence → select each clip → dub to target language → Replace. When complete, you have a fully localized sequence with all cuts, transitions, and timing intact. Repeat per language. The original sequence stays untouched as reference.

Limitations

Limitation	What it means
target_lang is a text field	No dropdown — you must know the ISO 639-1 code. Typos fail silently
No live cost estimate	Cost badge shows base rate only — check fal.ai for per-minute pricing
Replace overwrites original	No option to swap audio only — the dubbed mp4 replaces the full clip
Translation quality varies	Always review before delivery — especially legal copy, brand terms, humor
Lip-sync most fragile in close-ups	Test talking heads more carefully than wide shots
Busy source audio hurts results	Music and overlapping speech reduce transcription accuracy

Voice-over workflow with ElevenLabs — for creating or replacing narration from scratch.

Common failure modes — if generation fails or the dubbed result sounds wrong.