How Is Game Voice-Over Recorded? Five Recording Types and Lip Sync

Game voice-over is the most expensive stage of localising a game and the one that takes the most people. Translating the text is only the first step; the recording studio, the sound director, the adapter and the voice actors come in after it. The decision is therefore as much a budgetary one as a linguistic one.

Contents

The book underlines an important distinction: sound recording in games is not of a single kind. Five separate recording types are used, according to whether the character is visible on screen, whether the scene is finished, and what the technical constraints are (Jiménez-Crespo 2024, p. 218).

In this article we take up those five types, why lip sync is a cost item of its own, and when subtitling is the reasonable choice instead of voice-over. For other localisation topics you can look at our knowledge centre.

How does game voice-over differ from film dubbing?

Dubbing is the replacement of the source soundtrack with a recording in the target language; the aim is for the viewer to believe that the characters are speaking in their own language (Díaz-Cintas and Orero 2010, cited in Jiménez-Crespo 2024, p. 212). The definition holds for games as well, but the practice differs.

The first difference is linearity. A film has a screenplay; a game has independent pieces of audio called up according to the player’s choices. In the book’s words, these pieces “do not form a traditional film script”; they are text strings in spreadsheets (Mejías-Climent 2022, cited on p. 211).

The second difference is the lack of standardisation. Film dubbing has settled norms; in the games industry, practice varies from studio to studio. For the party buying the translation this means both flexibility and uncertainty: most of the time you have to write the rules yourself at the start of the project.

The third difference is the position of the audience. In film the viewer is passive; in a game the player triggers the sound through their own actions. The same line may be heard hundreds of times, or never at all. This means that the style has to be built more neutrally in lines that will be heard again and again, and more densely in one-off cinematic lines.

Where does the voice-over text come from?

Audio assets are usually held in spreadsheets. Every audio item is a separate segment and carries its own string ID. The screenshot in the book shows the audio units listed together with their ID name and with the name of the voice actor who is to record them (Toftedahl 2021, cited in Jiménez-Crespo 2024, p. 215).

In these spreadsheets the audio strings often sit mixed in with the other game texts, but they are marked as a “voice-over asset”. The translator’s first task is to read that marker: one line in the same file will be printed on screen, another will be read into a microphone.

The book makes a striking observation: in games the subtitle text and the dubbing text are most of the time the same text, whereas in film and television they diverge (Jiménez-Crespo 2024, pp. 212–213). The price of this is that long lines written for speech become unreadable as subtitles.

Five recording types and their effect on the translation

O’Hagan and Mangiron’s classification divides game voice-over into five types. The distinction may look technical, but it shapes the translated text directly: in one type you can write freely, in another you come down to counting syllables.

Recording type	When it is used	Constraint it places on the translation
Wild	When no character is on screen	Almost none
Time-constrained	When the sound has to match the original length	Total duration
Sound sync	When it has to match the character on screen	Start and end points
Lip sync	When the mouth movements are visible	Syllables, open and closed sounds
Stitch	When fragments are to be joined afterwards	Grammatical flexibility

The stitch method is used to save memory. In the book’s example “Your score” is recorded once and fragments such as “five goals” are then added to it (Jiménez-Crespo 2024, p. 218). In Turkish this structure cannot be built as easily as in English, because of suffix agreement.

Time-constrained recording is a heading of its own. The book gathers under it the cases in which the sound has to match the original length for technical reasons, and adds that new techniques can stretch this constraint. Even so, on most projects the translator has to write a text that will stay within the seconds allowed.

Why is lip sync a separate cost item?

Lip sync requires the recording to coincide with the mouth movements of the character on screen. The book states that this process is vital above all in close-ups of the face or the mouth (Jiménez-Crespo 2024, p. 212). Where there is no close-up, the requirement falls away with it.

On the technology side the picture is changing. Motion capture and facial animation techniques make it possible to produce characters’ lip movements separately for more than one language; in the example of Cyberpunk 2077 the book reports that as many as ten languages can be reached (Mejías-Climent 2022, cited on p. 206).

Even so, not every production has this capability. On most projects the translator has to write a text that will fit the source animation. That means moving away from word-for-word translation towards a rewriting that preserves the meaning while settling onto the mouth movement.

The book also states that synchronisation can be achieved through artificial intelligence assisted development techniques (Jiménez-Crespo 2024, p. 212). Tools of this kind are changing quickly, so it is safer not to assume that a solution valid on one project will hold on the next, and to tie the contract to the output rather than to the technology.

Who is in the recording room?

The recording stage is not a job done on one’s own. The book states that the process runs with a team made up of professional voice actors, a sound engineer, an adapter and a dubbing director (Jiménez-Crespo 2024, p. 218). The translator is the person who produces this team’s text.

The adapter’s job is to make the translated text ready for recording: breath points, syllable count, stress. On some productions this role does not sit with a separate person and is passed to the translator. Setting that division down in the contract prevents the later argument that “this was not my job”.

The larger the team grows, the larger the cost of re-recording grows with it. The text therefore has to be approved before it goes into the studio. A terminology error noticed during the recording can mean setting the studio up again for a single line.

Voice-over or subtitles? Decision criteria

The book calls dubbing “the most costly and most detailed method of transferring audiovisual material into another language”; subtitling, by contrast, is fast and cheap (Jiménez-Crespo 2024, pp. 211–212). The decision is taken between these two poles.

The publisher may change its mind even after the process has started. The book states that the decision to move from subtitling to dubbing, or in the opposite direction, can be taken later on. This flexibility means the translated text has to be written from the outset to suit both scenarios.

In practice the criteria come under three headings: the visual density of the scenes, the size of the target market and the genre of the game. In an action game full of cinematics, voice-over improves the experience noticeably; in a text-heavy strategy game, subtitles are usually enough.

A middle way is possible as well: recording voice-over only for the main story cinematics and leaving the side quests to subtitles. This mixed model brings the budget down, but it carries a risk of inconsistency. If the player hears Turkish in one scene and then returns to English in the next, the illusion can break, so the transitions have to be planned.

Audio description: the accessibility layer of sound

In recent years some productions have offered audio description for visually impaired players. The book gives the example of The Last of Us Part I and defines this as a form of “assistive audiovisual translation” (Perego 2019, cited in Jiménez-Crespo 2024, p. 213).

Audio description calls for a team of its own: describers, voice actors, sound technicians and, preferably, visually impaired consultants. The language used is expected to be “meticulous, concise, visually dense and usable”.

The book says that the practice is still rare in the industry, but that it is expected to spread. Putting it on the table as an option now, while the localisation budget is being drawn up, comes cheaper than adding it afterwards.

The audio description text has to use the same term list as the game’s other audio assets. Otherwise the same object is called one thing in the description and another in the dialogue. Accessibility features also have to be verified at the testing stage; the book states that this is not yet common practice in the industry (Jiménez-Crespo 2024, p. 220).

The Türkiye context

In Türkiye, Turkish game voice-over is seen far more rarely than subtitling. Common practice is to put the menus and the subtitles into Turkish and leave the sound in English. Even so, when some large productions are released with Turkish voice-over, they meet a marked response in the player community.

The structure of the language is a factor too. Because Turkish is agglutinative, the predicate of the sentence falls at the end; where an English line completes its meaning early, the Turkish equivalent closes late. In time-constrained and lip-synced recordings this calls for the sentence to be rebuilt.

In stitch recordings the problem is sharper still. The “Your score” + “five goals” pattern may not sound natural in Turkish because of suffix agreement. The solution is usually to redesign the fragments so that they carry no suffix, or to settle on a fixed pattern. This is a design decision rather than a translation decision, and it is taken together with the developer.

What does this mean for you?

If you are setting aside an audio budget for your game, the sentence “we will have voice-over done” is not a description of scope on its own. Which recording type is used in which scene changes the cost several times over. The list below sums up what to do before the decision.

Separate the audio strings from the other texts and label the recording type of every line.
List the scenes that contain close-ups; apply lip sync only to that list.
Add a maximum length column to the lines that carry a duration constraint.
Review the stitch structures separately for Turkish; if suffix agreement breaks down, change the pattern.
Plan the subtitle version as a separate deliverable and do not use the recording text as it stands.
Have the text approved together with the term list before it goes into the studio.

These steps reduce both studio time and the risk of re-recording. If you would like to talk through how the audio assets in your project should be handled, you can share your file structure through our quote form and look over our services page for scope options.

Frequently asked questions

How many types of recording are there in game voice-over?

The literature counts five: wild recording, time-constrained recording, sound-synced recording, lip-synced recording and stitch recordings. Which of them is used depends on whether the character in the scene is visible on screen.

Is lip sync needed in every scene?

No. Lip sync matters mainly in scenes where the character’s face or mouth is brought into close-up. For narration in which the character is not on screen, wild recording is enough and much cheaper.

Are the subtitle text and the voice-over text the same in games?

Usually yes. The translated audio strings in a game are used both in the subtitles and in the recording. This can lower subtitle readability; in film and television the two texts are prepared separately.

What is stitch recording for?

Stitch recordings are fragmented recordings that will be joined together afterwards. “Your score”, for example, is recorded once and different number fragments are added to it. It saves memory, but in Turkish it has to be constructed carefully because of suffix agreement.

Sources

Jiménez-Crespo, M. A. (2024). Localization in Translation. Routledge, pp. 206–218.
Mejías-Climent, L. (2022). Enhancing Video Game Localization Through Dubbing. Palgrave Macmillan.
Perego, E. (2019). “Audio Description”, The Routledge Handbook of Audiovisual Translation.