Forced subtitles vs full subtitles: what belongs in each?

Forced subtitles cover selected moments

Forced subtitles supply selected text that viewers need even when they are not following a full subtitle track. That might be the translation of a brief conversation in another language or a written message that matters to the story. A full subtitle track covers the dialogue throughout; accessibility captions also convey relevant sounds and speaker information. A sparse forced track cannot replace that complete coverage.

The word “forced” describes intended playback behavior, but selecting the right content is an editorial job first. Start with the audience and the audio version, then decide which moments need text. Only afterward configure the receiving player or delivery package. This guide uses an original fictional short film; apply the workflow to media and translations you own or have permission to edit and distribute.

Name the audio version and intended audience

Write down the final picture edit, selected audio language, subtitle language and destination. “English subtitles” is insufficient: the file might accompany English original dialogue with a brief French exchange, or an English dub that already translates that exchange. The necessary text can differ even though both audio tracks are labeled English.

Watch the exact audio version from beginning to end. For each candidate moment, ask whether the intended viewer needs the information, whether the audio already supplies it, and whether the picture already makes it readable in that viewer’s language. Do not automatically translate every visible sign or unfamiliar utterance. If a filmmaker deliberately withholds a message from the audience, confirm that intent before revealing it in subtitles.

Netflix’s forced-narrative guidance includes translated dialogue and on-screen communication among its use cases. Its delivery rules are a useful example of an explicit contract; they are not a universal specification for every video service. Ask the recipient which events and file types their delivery actually requires.

Make an inclusion ledger before cutting a file

Keep a small review note with the time, what appears or sounds, the proposed wording, and the reason to include or exclude it. Here is an invented example for a short film with mostly English dialogue, intended for an English-speaking audience. All times and scene details are illustrative.

At 00:12, a character says “The workshop is closed” in English. Keep it in the complete captions; it needs no separate translation for this audience. At 00:28, another character says “Attends ici” in French, and the viewer is meant to understand the instruction. The English rendering “Wait here” belongs in the essential subset and the complete English track.

At 00:41, a close-up shows a French note saying “Retour à midi,” with no spoken explanation. The approved English text is “Back at noon.” Include that information in the subset and account for it in the complete English delivery. At 00:55, an off-screen knock changes the character’s reaction. Describe the knock in the accessibility captions; it does not belong in this translation-only subset just because it matters.

Record unresolved meanings or creative decisions as pending. A timecode and a guess are not an approved cue. W3C’s caption guidance explains why speech and meaningful non-speech audio both matter; the full accessibility track needs its own review beyond these selected translations.

Keep the full track complete

Preserve the approved complete file. Build a separate working copy for the selected events, using an editor that supports the necessary cue selection and authoring. Keep each event at its original position in the final film. A sparse file should contain long gaps where no text is needed; do not move the remaining captions together to close those gaps.

Record a mapping from each selected event to its place in the full track. If a written message never had a cue, author and time it against the picture in a suitable editor. Do not attach it to the nearest spoken sentence merely because that sentence already has timestamps.

Netflix specifically requires forced-narrative events in its full subtitle and SDH/CC files because its service stops displaying the separate forced track when a full track is enabled. For another destination, confirm the equivalent switching behavior. The practical acceptance question is whether viewers still receive every essential translation after selecting full captions, without seeing it twice.

An SRT filename does not force playback

The two approved events from the fictional ledger could produce the following SRT excerpt. Its numbering starts at one, but its times still refer to the original film. The file contains only timed text; this example does not package or enable a track in a player.

A descriptive name such as workshop.en.essential.srt helps a person identify the deliverable. Some playback systems recognize their own filename conventions, but adding “forced” to a filename is not a portable command. Confirm the recipient’s exact convention or packaging requirements. A regular SRT file has no standardized forced-track switch that will make every player display it.

1
00:00:28,000 --> 00:00:30,000
Wait here.

2
00:00:41,000 --> 00:00:43,500
Back at noon.

Separate selection, track flags and burned-in text

Three different decisions are often hidden behind one checkbox: which events a file contains, when the player selects that track, and whether text is permanently written into the picture. Keep those decisions separate in your handoff.

Matroska documents different Default and Forced track flags. Default is a hint about automatic track selection; Forced signals that a subtitle track should display even when ordinary subtitle preferences would not select one. Setting either flag does not translate dialogue or remove unwanted cues. The player and its settings still matter, so verify the final file in the actual destination.

HandBrake’s documentation describes sources with a separate forced track and sources with marked events inside a larger track. Its documented Foreign Audio Search works with bitmap subtitle sources; it is not a promise to recognize foreign-language sentences in an arbitrary SRT. Read what a tool’s option applies to before using it on your file.

Burned-in text is a separate delivery choice: it becomes part of the video image and cannot be switched off. Check the clean master before adding another track, because an existing translation in the picture can collide with a second copy. Retain a clean source if you control the production.

Test the viewing states that can lose information

After the destination’s packaging or upload step, test the intended audio with full subtitles disabled. Confirm that every approved essential event appears, ordinary dialogue stays absent, and the first and last selected events have correct timing. Seeking directly to each cue is useful, but also play into it from the preceding scene.

Next enable the complete subtitle or accessibility track. Check that the same essential information remains available once, with no doubled translation or collision with on-screen text. Then change to each supported audio version and repeat the affected moments. A dub may change what needs translating; do not assume one sparse file is correct for every soundtrack.

Finally check the finished deliverable on the intended screen and player. Record the audio selection, subtitle setting, player or platform, video revision and observed result. A local preview establishes only that local behavior. If the receiving service has not been checked, label destination acceptance pending rather than reporting the file ready.

Where Polyglot fits in the handoff

Polyglot can help create, translate and review timed subtitle text locally on an Apple Silicon Mac, then export separate SRT or WebVTT files. Review the wording against the recording and keep an untouched complete export before preparing any selective derivative. AI output needs human review, especially when a short translation carries a plot detail.

Polyglot does not configure forced-track flags, package subtitle tracks into a video container, burn captions into the picture or decide which narrative information viewers should receive. Use the destination’s supported tools for that delivery stage. Send the complete track, approved essential subset, inclusion ledger and playback notes together, with the intended audio version named clearly.

Keep exploring.