Actual-play subtitles. Keep the ruling in the story.

Subtitle the session that was played

In a recorded tabletop role-playing session, “I open the door” may be a proposed action. A roll, interruption or ruling can change what follows. Captions that omit “I try” or a spoken correction can turn an attempt into a completed event. Review the exchange as a sequence before you approve the individual lines.

Polyglot can create timed subtitles from your authorized recording, translate reviewed text locally on an Apple Silicon Mac, and export separate SRT or WebVTT tracks. This workflow is for the finished video or audio episode. It does not provide live table captions, recognize speakers, interpret dice, enforce game rules or write a session recap.

The official D&D playing guide distinguishes scene description, player actions and narrated results. That is one system’s example of why the surrounding exchange matters. Use the actual group’s recorded words and decisions; do not replace them with what a rulebook says should have happened.

Establish the release boundary and episode references

Start with your own production or a recording you have permission to process, translate and distribute. Confirm the intended audience with participants and check included music, artwork, game material and guest contributions. Another group’s publicly watchable episode is not permission to download, translate or republish it.

Finish cuts, breaks and any removed conversation in the media editor before generating the delivery track. Keep the episode identifier, final media filename and source language with your review notes. If a private aside is removed, check that it is also absent from the exported captions and transcript. Polyglot does not manage consent or control access to shared files.

Ask the episode owner for spellings of names and places already introduced, the rules edition used and any relevant table conventions. Keep unrevealed identities and private campaign notes separate from the public subtitle handoff. Reference notes help a reviewer understand a line; they are not additional dialogue to insert.

Review the attempt, roll and ruling as one exchange

Make a small external review sheet for exchanges where the first statement is conditional, corrected or superseded. Record the media range, proposed action, audible roll or total, final ruling and any unresolved wording. This is a manual editorial aid, not a feature that Polyglot generates.

Play through the resolution, not just the first cue. Preserve a player’s change of mind and the game master’s correction where they remain in the final cut. A result announced as “twelve—sorry, fourteen” should not become a confidently spoken twelve, nor should the earlier words disappear without an editorial reason. Do not calculate a replacement total from a visible die or silently repair a ruling.

The fictional example below shows the review distinction. The sheet summarizes the exchange for the editor; the subtitles themselves follow the audible words. “Attempt fails” is a review note, not an extra caption.

Fictional episode
lantern-07-v3

12:04 Player
“I try the latch.”
12:08 Player
“Twelve—no, fourteen.”
12:12 GM
“It still will not open.”

Review decision:
Keep “try” and the correction.
Keep “not” in the ruling.
Do not caption:
“The door opens.”

State: attempt fails
Approval: owner + reviewer

Separate table discussion from fictional speech

A participant can describe an action, speak as a character and ask a rules question in the same minute. Review ambiguous transitions with the episode owner. A hypothetical “would that wake the guard?” is not a statement that the guard woke up. Likewise, an out-of-character joke need not become something a character said inside the story.

Use a consistent, restrained identification convention when the listener would otherwise lose track. DCMP’s guidance distinguishes performers from portrayed characters and advises against naming someone before the audio or picture introduces them. Polyglot does not decide those identities. Manually review any plain-text labels already present or added to an existing cue; special positioning and styling require separate destination checks.

Critical Role’s FAQ describes an editorial captioning process that pays attention to names and lore. That supports the need for review, not a claim that Polyglot is used by or endorsed by the show. For your own episode, keep a short approved terminology list and mark genuinely unclear words for a human decision rather than guessing a familiar character name.

Carry the same decisions into every language

Review the original wording before translation. Give the language reviewer the accepted attempt-roll-ruling sheet and only the audience-appropriate terminology references. Check conditionals, negation, questions and corrections: “I could” is different from “I do,” and a spoken total is different from the number visible on a die.

Keep game terms consistent where they refer to the same thing, but distinguish ordinary speech from a rules term. A participant may use “save” in casual conversation as well as in a game mechanic. Resolve ambiguous usage from the exchange rather than applying a glossary mechanically. Polyglot does not adjudicate rules or validate fictional continuity.

Translation preserves source timing during processing; you can explicitly adjust cue timing for a selected language. Review each language in playback so a ruling does not appear before it is spoken. If overlapping speech needs separate new cues, or essential sounds need their own timed descriptions, finish that work in an editor that supports individual cue creation. An automatic speech track alone is not a complete caption review.

Accept the final episode and every sidecar file

Export one reviewed file per language, named with the exact final episode identifier. For YouTube, the documented workflow allows a supported subtitle file to be uploaded with timing. Confirm the receiving player’s format and language requirements. Polyglot does not upload the files, burn captions into the picture or replace a track already published elsewhere.

Check the opening and final speech, then every flagged attempt-to-ruling exchange and transition between table talk and fictional dialogue. Verify names, spoken corrections, meaningful sounds, narrow-screen wrapping and that cues clear through breaks. These checks supplement a full review of the track; spot checks alone do not establish accuracy or accessibility.

If the episode cut changes, review the affected sequence again and replace all affected deliverables through your publishing process. Keep old and new versions distinguishable. A readable transcript can accompany the episode, but it is not an automatically generated list of canonical events or a substitute for the timed subtitle track.

Polyglot requires Apple Silicon and macOS 14.4 or later. Media and text processing stay local; requested downloads, user-controlled updates, default-on content-free analytics with an opt-out and Pro licensing have separate network boundaries. New AI subtitle jobs use a seven-day trial and require Pro afterward; editing and exports remain free.

Keep exploring.