Commentary subtitles. Match the voice viewers choose.

One picture can have two different conversations

A film’s ordinary dialogue captions do not describe a director talking over that film. Even when both soundtracks are English and have the same duration, their words and useful sound information differ. Create and review a separate caption track for the commentary your audience will actually hear.

Polyglot can transcribe an authorized commentary reference locally on an Apple Silicon Mac, let you correct its timed text and export SRT or WebVTT. The film editor or delivery owner handles audio mixing, track packaging and the destination player. Polyglot does not author a disc, embed alternate soundtracks or automatically bind a caption file to an audio choice.

Import a reference that contains the intended soundtrack

Ask the editor for a supported audio file or review video containing only the intended final commentary mix as its audio track. Listen before importing. A movie containing several embedded soundtracks is not a reliable way to choose the commentary: Polyglot’s current extraction reads the first audio track. Selecting commentary in a different player does not change what Polyglot imports.

Keep the reference on the same timeline as the delivered film, including its opening silence and any lead-in. Record the film cut, commentary mix revision, source filename and clock origin. Check a recognizable passage near the beginning and another near the end against the delivered version. A separately recorded conversation that starts later needs alignment in the editing workflow before you rely on its caption times.

Transcribe the verified reference, then listen while reviewing the words. When the commentary mix also contains audible film dialogue, decide what needs representing in that mix. A clean microphone-only recording can help resolve a word, but it is not a substitute for reviewing the final mixed soundtrack. Polyglot does not separate overlapping voices or automatically identify the speakers.

Review commentary as commentary

Check names, locations and references to the production carefully. Distinguish a commentator quoting a character from the character speaking in the underlying film. Add a speaker label only when you can verify who is speaking and the viewer needs that information. Keep uncertainty visible during review instead of turning a plausible name into an approved fact.

Include meaningful audible information needed to follow this soundtrack, and check that captions clear when they should. W3C guidance treats accurate speech and relevant non-speech audio as part of caption preparation; automatic transcription alone is a starting point. For dense overlaps, have a knowledgeable reviewer decide how to represent the audible exchange clearly. Do not combine two complete caption files and assume the result is readable.

If a meaningful event has no timed cue, finish that interval in a caption editor that supports adding and timing cues, then import the reviewed SRT or WebVTT. Do not stretch unrelated dialogue to cover a missing sound. Polyglot does not promise automatic sound-event caption creation.

Translate only the reviewed commentary track when another audience language is needed. Keep the main-film translation separate. Matching languages do not make the two source tracks interchangeable, and a fluent translation still needs a person to check production terminology, jokes and references.

Name the role as well as the language

Use a small handoff map outside Polyglot to identify each approved audio/caption pair. Include the soundtrack role, spoken language, caption language, film revision, caption filename and intended viewer-facing label. Two menu entries called only English leave the important distinction hidden. The delivery owner must check which labels the destination actually exposes; a filename alone does not set every player’s menu label.

This invented example uses one film cut with two English soundtracks and two independent caption files. The map is a production note, not an app-generated playlist or a guarantee that a player will link the tracks automatically.

Fictional film: Cut 04
Both mixes: film origin 0

Main mix / English
Caption: cut04-main-en.srt
Label: English — Film

Commentary mix / English
Caption: cut04-notes-en.srt
Label: English — Commentary

At 00:02:14:
Main: “Leave the key.”
Notes: “We shot this twice.”

Check audio AND text.
Same time, different words.

Test the pair in the receiving player

VideoLAN’s VLC desktop documentation describes audio-track selection, subtitle-track selection and loading an SRT file. Those are separate controls to inspect, not evidence of automatic soundtrack/caption pairing. Follow the controls in the installed version and confirm the actual media package works. A check in VLC also does not establish behavior in a website, television app or disc player.

Start with a fresh opening of the final package. Confirm the default audio and caption choice, then select the commentary audio and its captions. At the example’s shared timestamp, the visible text must match the selected voice. Seek forward, pause and resume, then switch back to the main soundtrack and its captions. Check that an old cue does not remain on screen and that the commentary file has not become the main film’s default.

Repeat after closing and reopening the media, and test each offered caption language with its intended soundtrack. Record the player and version, selected labels, tested passages and any manual selection viewers must make. If the destination cannot distinguish or reliably deliver the pairs, resolve the packaging there. Separately labeled main-film and commentary videos, each with its own reviewed captions, may be a clearer delivery choice.

After a commentary remix or film recut, treat the affected pair as needing review again. Compare the new reference, replace only its intended caption asset and repeat the switching checks. Keep earlier approved files available until the replacement is accepted; a successful file upload is not proof that the audience sees the right captions.

Keep the release permission and product boundaries clear

Work from films, recordings and commentary contributions you own or have permission to process and distribute. Confirm the agreed scope for the commentary, quoted material, translations and audience access. Recording-session chatter or an unreleased detail may be audible but not approved for publication. Keep those decisions with the delivery owner and correct the source release before publishing its captions.

Polyglot requires Apple Silicon and macOS 14.4 or later. Recognition supports 25 European languages; translation targets are a separate capability. Media and text processing stays local after model setup; the privacy page explains downloads, licensing, updates and analytics. New AI jobs are unlimited during the seven-day trial and require Pro afterward; editing and exporting completed tracks remain free.

This is a manual preparation and acceptance workflow. It does not claim tested compatibility with your delivery platform, automatic track selection, audio separation or accessibility certification. Release the commentary only after a reviewer checks the final sound, text and player choices together.

Keep exploring.