Explore Polyglot

Download now

For Apple Silicon · macOS 14.4+

Local AI. Your Mac.

Made for your workflow

Game cutscenes. Keep subtitles on the scene clock.

Create reviewed timed text for a fixed cinematic, then hand it to the developer with scene IDs, language files and playback checks for the actual game build.

Download now

Apple Silicon · macOS 14.4 or later

01

Identify the cinematic

Bind the finished media and each language track to a stable scene and revision.

02

Review the dialogue

Check character names, delivery, meaningful sounds and the clip’s own timing.

03

Test the game handoff

Have the developer verify subtitles during normal play, pause, skip and restart.

A black iridescent foil clapperboard with a play symbol, a directional pad and a floating subtitle tile.
AI-generated editorial illustration

Prepare the text for a fixed cinematic

A subtitle can be perfectly timed in a video player and still appear at the wrong moment inside a game. A cinematic might start after a loading screen, pause while the player opens a menu or disappear when they skip ahead. Its text needs to follow that cinematic’s playback, not the time since the game launched.

Polyglot helps you create, review and translate timed subtitles from an authorized prerecorded cutscene on an Apple Silicon Mac running macOS 14.4 or later. Export SRT, WebVTT or text for the developer’s agreed handoff. The developer implements and tests subtitle display in the game; Polyglot does not import into an engine, create localization keys or control runtime captions.

This workflow is for a fixed audio/video sequence. Branching conversations, variable dialogue order, player names inserted into speech and gameplay-triggered voice lines need a game-specific dialogue system. A recording of one playthrough cannot establish the timings for every possible playthrough.

Agree on the receiving format before processing

Ask the developer which format their subtitle pipeline actually accepts, how it identifies a scene, which clock drives playback and how it selects a language. SRT or WebVTT export alone does not prove engine compatibility. If conversion into a string table or custom data asset is required, assign that conversion and its validation to the developer before making a batch of files.

Keep a manual manifest with a stable cinematic ID, exact media filename and revision, audio language, subtitle language, duration and receiving build. Cue numbers are positions in one subtitle file, not durable game localization IDs. If an external tool renumbers, combines or splits cues, preserve a mapping and review any lost or duplicated lines.

Use the final picture and audio mix. Remove a production slate or countdown in the video editor before transcription, or document its intended treatment with the developer. A trailer trimmed from the cinematic is a separate edit and needs separate timings. A new dub may have different pacing even when the pictures have not changed.

Keep the scene clock separate from the game clock

Time the subtitle file against the beginning of its matching media file. Do not add the menu wait, loading delay or the time a player spent reaching the scene. Those values vary between sessions and belong to playback implementation, not the caption file.

Unity’s VideoPlayer documentation distinguishes the current video time from its clock source, and exposes pause, stop, seek-completion and end-of-playback behavior. These controls show why the developer needs an explicit timing contract. They are not evidence of a Polyglot engine integration or universal sidecar support.

This fictional handoff note illustrates the contract. It is a manually prepared production note, not an app-generated report or a tested game integration. The line belongs at 12 seconds into this video regardless of when the player starts it.

Fictional handoff note
Scene: observatory_intro
Media: obs-intro-r04.mp4
Audio: English
Track: obs-intro-r04.en.srt
Clock: media begins at 0
Cue 7: 00:12.000–00:14.400
Mira: “Leave the lens here.”
Skip: clear scene captions
Restart: start from media 0

Review the performance, not only the script

Import the finished media into Polyglot and create a draft in a supported recognition language, or import an existing reviewed SRT or WebVTT. Listen beside playback. The performed dialogue may differ from the screenplay: an actor may change a name, repeat a phrase, interrupt another character or leave a sentence unfinished. Preserve what the final recording actually says.

Keep a small human reference sheet for invented place names, character names, item names and established translations. Check whether a voice is offscreen, over a radio or deliberately unidentified. Avoid revealing a character’s identity in the text before the story reveals it. Polyglot does not automatically identify speakers, recognize fictional terminology reliably or infer story context.

Microsoft’s Xbox Accessibility Guideline 104 covers subtitles for speech and captions for important non-speech audio, including cinematics. A speech draft is not a complete accessibility track. Review meaningful alarms, footsteps, music changes and other sounds; add or finish missing captions with a suitable editor where needed. Review placement and readability in the game separately, and do not treat this checklist as certification.

Translate against the intended voice and story

Review the source before translating. Preserve a threat, a question, a joke or a hesitant answer rather than turning every line into neutral instruction. Use established names consistently, keep clues intact and check that a translation does not disclose information the scene intentionally withholds. A fluent reviewer should compare the translated text with the performance and story context.

Subtitles translated for the same audio can begin from its existing timing. A separately dubbed audio track needs another timing review against that exact dub; source timestamps cannot guarantee alignment to a new performance. Keep each audio/subtitle pairing explicit in the manifest. Do not assume a language dropdown selects the intended combination.

Confirm permission to process and translate the recording, including voice performances, licensed footage and unreleased material. Follow the team’s AI-use and distribution rules. Local processing does not grant rights or enforce an embargo. Polyglot keeps media and transcript processing local; requested downloads, user-controlled updates, content-free analytics with an opt-out and Pro licensing have separate network boundaries.

Accept the result inside the playable build

First reopen each exported track with its exact media and check the first line, a distinctive middle exchange, the final line and silent gaps. Verify expected cue count, order, text, accents and any approved timing changes. Give the developer the reviewed files, manifest and unresolved holds; preserve the original exports beside any converted engine assets.

Then test the actual target build. Start the cinematic after a short wait and after a longer wait. Pause and resume if supported. Skip during a visible line and confirm it clears. Restart the cinematic or reload its entry point and confirm no old line remains. If seeking or looping is offered, verify the text follows the new media position. If these controls are unavailable, record that limitation instead of claiming they passed.

Check every supported audio/subtitle pairing and the project’s intended language-change behavior. Watch with sound off, inspect the largest subtitle size and confirm text does not cover essential action or game controls. A missing glyph, stale line after a skip, mismatched dub or track from the wrong revision is a delivery hold. Loading a file successfully is only the start of acceptance.

Keep the tested build identifier with the approved media and language files. These checks are a suggested developer acceptance procedure, not tests performed on your game by Polyglot. For a developer update published as a regular video, use the separate game-devlog workflow; its destination is a video player rather than runtime game playback.

Sources & further reading

Microsoft Xbox Accessibility Guideline 104: subtitles and captions

Unity: VideoPlayer timing and playback controls

W3C: transcribing audio and reviewing automatic drafts

Polyglot: local processing and network boundaries

Published by Polyglot using an AI-assisted editorial workflow. How we prepare and update our guides.

Keep exploring.

Try it on your Mac.

Find the right tools for your words and video, with processing on your Mac.

Download now

Apple Silicon · macOS 14.4 or later