A perfectly timed English line can unravel the moment it crosses into another language. The character’s mouth keeps moving after the audio ends, or the delivery races so the actor sounds like they’re racing through a checklist. Players notice. Immersion breaks. Reviews mention “awkward dubbing.” That friction sits at the center of game voice-over work: timing constraints that force every translated line to land inside a fixed window set by animation, facial performance, or engine triggers.
Languages do not expand or contract at the same rate. English often favors concision. German and Russian versions commonly run 20–30 percent longer. Japanese and Chinese can pack dense meaning into fewer syllables, yet the spoken rhythm, pauses, and intonation still shift the actual duration. When the source audio is locked to visible mouth shapes or gesture cues, the target performance has no room to wander. The result is familiar: rushed delivery that feels like chanting, or stretched lines that leave the lips flapping in silence.
Industry practice sorts these limits into practical tiers. Wild dialogue—off-screen narration or ambient NPCs—lets actors speak at natural length. Loose constraints allow roughly ±20 percent. Soft ones tighten to ±10 percent. Strict timing demands near-exact duration. Sound-sync requires matching the original’s internal pauses. Full lip-sync goes further: the words themselves must align with the visible articulations. Recording rates drop accordingly. Wild lines can move at around 100 per hour. Lip-sync work often slows to 10–15 lines per hour because directors and actors adjust delivery syllable by syllable. Cinematics absorb most of the strictest requirements; gameplay barks usually stay more flexible. Treating every line as lip-sync wastes budget. Ignoring it on close-up emotional beats produces the mismatches players call out.
Teams that handle this well treat timing as a design constraint from the first translation pass, not a post-recording fix. Translators receive timing notes or character limits alongside the script. They adapt rather than translate literally—shortening expansive constructions in German or Spanish, expanding sparse ones where needed, while preserving emotional intent and character voice. Placeholder English audio frequently serves as a shared reference track so every language works to the same beat. Collaboration between translators, voice directors, and narrative leads flags problem lines early. When animation can still be adjusted, some studios stretch or compress the visual sequence slightly rather than force unnatural speech rates. For locked cinematics the script must flex.
Real projects show the stakes. CD Projekt Red’s work on Cyberpunk 2077 involved nearly 2,500 people in localization alone, including roughly 2,000 voice actors across 11 languages for tens of thousands of lines. They used a combination of machine learning and rule-based systems to generate lip-sync for multiple languages rather than scaling English-based facial animation. The goal was simple: a character speaking Mandarin should look as if they are speaking Mandarin, not a stretched English performance. Earlier titles such as The Witcher 3 had already tested algorithmic facial work; Cyberpunk pushed it further so more languages could be treated as first-class. By contrast, early overseas feedback on Black Myth: Wukong highlighted mismatches between English voice tracks and Chinese lip movements, underscoring how siloed pipelines create visible problems that players immediately notice.
Perception research reinforces the technical thresholds. At standard frame rates an average audiovisual offset beyond roughly 100 milliseconds becomes distracting; ideal alignment stays under 45 milliseconds—about one frame. Once viewers register the disconnect, the scene stops feeling native. Market data adds weight: voice-over and dubbing form a growing share of game localization spend, driven by narrative titles and player preference for fully voiced experiences in key territories. Text localization alone no longer suffices for immersion-heavy games.
The practical path stays consistent across studio sizes. Provide timing references early. Classify lines by constraint type so budgets and schedules match reality. Record against visual or audio placeholders. Allow iterative passes—meaning first, then timing, then performance. Accept that some languages will need creative rephrasing approved by the narrative team. Hybrid tools can accelerate alignment and rough passes, yet the final emotional accuracy still rests with native performers who understand both the language and the character.
Artlangs Translation has spent more than two decades refining exactly these workflows across more than 230 languages, drawing on a network of over 20,000 professional translators and a track record of successful game, video, short-drama, and audiobook projects. Their teams routinely handle the full chain from adaptive script work through multilingual voice-over, subtitle localization, and data annotation, giving studios a single partner capable of keeping both meaning and mouth movements in step.
