The Master Guide / 10 Voice and lip sync · Speaking shots and lip sync
Choosing the route per shot
Choose each speaking shot's route in your shot plan (the plan you make before generating; see Planning, budgets and takes), before anything is generated.
Choose each speaking shot's route in your shot plan (the plan you make before generating; see Planning, budgets and takes), before anything is generated.
Two terms run through this section. A sound-off speaking plate is a Seedance shot of the actor speaking the line with the sound switched off: the mouth moves at the recording's measured beats, and no voice comes back. Sync-3 (Sync Lipsync 3, run on Higgsfield) is the lip-sync model that then fits that mouth to the exact recording.
Shot | Route | Label |
Any Arabic speaking shot, medium close-up or closer, the mouth readable at delivery size | Route 2, the default: a sound-off plate of the actor speaking comes first; Sync-3 then fits it to the exact approved recording | (decided) |
A close-up hero line where body and breath must ride the voice (a laugh breaking into the line, a sob, a shout, a punchline's timing) | Route 1, only if you choose it for this line: the audio-driven plate, where Seedance hears the recording. Every other Arabic close shot stays on Route 2 | an option [single strong source] |
A line the AI voice can't carry after the fix ladder (see Voice and Arabic dialects → Spelling for the engine); a project shot with real actors | A native actor's read driving the face | a fallback only (decided); the method [single strong source] |
Wides, small faces, backs, faces turned away | No lip sync; the line laid over the picture | — |
Songs and jingles sung on screen | Sync-3 on the isolated vocal stem | — |
Two or more speakers | Each generation and each Sync pass holds one speaker | [official guidance] |
A voiceover from off screen | No lip sync; on-screen mouths closed | — |
One-line extras, drafts, animatics, social cut-downs, outside Arabic | The re-voiced read, only where you accept it. No Arabic re-voiced read ships in any version: every Arabic voice is recorded | (decided) |
English principal lines | The recording as an audio reference, laid back, lips checked, Sync-3 only where needed | (decided) |
English, or another language Seedance speaks well | Seedance's own voice, from a client-approved sample | an option [proven] |
Route 2 is the Arabic default, a punchline or short line included. No route replaces another by itself, and a change of route is your choice for that shot, written in the plan (decided).
Route 1 is an option for close-up hero lines, named in the plan, priced, and made only when you choose it for that line, in the plan (see Speaking shots and lip sync → Route 1, for close-up hero lines). It is also the one alternative for plates that fail the first-plates check twice, again only when you choose it for those shots (see Speaking shots and lip sync → Route 2, the Arabic default).
Arabic is never laid back over a clip without a Sync pass, and Seedance never voices Arabic (decided): not its own voice (a voice Seedance makes from a written description), not a re-voiced read, not extras or a crowd murmur. Every Arabic voice is recorded.
Every speaking shot stays internal until the rough cut; the client sees speaking shots only inside it. The one exception is the voice sample for Seedance's own voice, which the client approves in the anchors presentation (see Speaking shots and lip sync → English lines and Seedance's own voice).
The whole lip-sync method is [best available, not yet validated]: it follows the tool makers and feature-film visual dubbing, but nobody has published this exact chain of Seedance 2.5, Sync-3 and Arabic, and Sync-3 on Seedance plates rests on one shipped creator case. Never present it to a client as proven.
Why this default: Sync's own documentation says its models keep the performance of the face they are given and need a face that is already speaking. Feature-film visual dubbing (Flawless's TrueSync on Watch the Skies) keeps the performed original and refits only the mouth. And no placing film at Higgsfield's 2026 festival shows Seedance speaking Arabic. The one practitioner reel with Egyptian lines (the Franco quote, Speaking shots and lip sync → Route 1, for close-up hero lines) hasn't yet been heard with the sound it ships. Seedance's success in Japanese (Fukushima, Soulless) is not evidence for Arabic.
Sources
Sync, lip-sync models and improving lip-sync quality; Flawless TrueSync on Watch the Skies (Decrypt); Higgsfield's Lipsync Studio post; Fukushima by @aist (project); Soulless by Yoav Yariv (project).