Workout video subtitles. Keep every round in sync.

Caption the workout viewers will actually follow

A follow-along workout can show a movement before the instructor asks the viewer to begin. Later, the same movement may return with a different countdown, an optional variation or a longer recovery. A caption copied from the demonstration or an earlier round can therefore be readable and still belong to the wrong moment.

Polyglot can transcribe your authorized recording locally on an Apple Silicon Mac, help you review and translate its timed text, and export SRT or WebVTT. Use the finished edit as the source. This workflow focuses on the boundary between demonstrating, doing and resting; it does not design a workout, assess movement or generate an exercise timer.

Use only recordings you own or have permission to process and distribute. Confirm permission for the instructor, visible participants, music and any included footage, including the intended translated publication. A subscription to someone else’s fitness library does not itself establish that permission. Keep a qualified content owner responsible for the exercise instructions; subtitle review cannot establish their suitability for a viewer.

Make a block map before reviewing individual lines

In a separate document, name the final video revision and mark the beginning and end of each demonstration, active round, rest and transition. Give repeated rounds their own identifiers. This is a manual production note, not an automatically generated Polyglot report.

Use the recording’s full playback clock. A graphic that resets to zero at the beginning of each round is a different clock. Do not put that round-relative time into an SRT timestamp. If an introduction or rest passage changes, recheck all later boundaries against the new final cut.

The fictional timing note below illustrates a caption review decision. The durations and words are examples, not a recommended workout or results from an app test. In particular, the spoken “three” should not appear during the earlier demonstration or remain visible after the active block begins.

Fictional final-cut review
Video: studio-rounds-v04.mp4

00:42–00:50  Demonstration
00:50–00:53  Spoken countdown
00:53  Active round A begins
01:23        Rest begins

At 00:50: “Three, two, one.”
At 00:53: “Start.”
At 01:23: “And rest.”

Use full-video cue times.
Check each phrase
against the audio.
Round B needs its own review.

Keep a spoken countdown attached to its sound

Listen across each start and stop boundary while watching the subtitle. Check missing numbers, repeated words and commands that recognition has joined into one long cue. A phrase such as “three more” counts remaining repetitions; it is not necessarily a three-second countdown. Preserve what the instructor says rather than converting every number into a timer.

Review both cue edges. A late “start” or an old instruction lingering into “rest” can misrepresent the current passage. Adjust existing cue timing in Polyglot for the selected language, then play the whole transition again. If a long cue needs splitting or a sound needs a new cue, finish that structure in a suitable caption editor and validate the returned file.

If the spoken countdown, the visible timer and the accepted edit disagree, flag the recording for the producer. Do not change a caption’s number to match a graphic while leaving contradictory speech underneath. Correct recognition to the audio; resolve an actual production mismatch in the recording before approving its caption track. Polyglot does not detect workout blocks, reconcile timers or judge the intended pace.

Review alternatives and repeated rounds separately

Keep the scope of an option intact. “You can stay with the first version” differs from “stay with the first version.” Check which movement the speaker is referring to and retain the words that make a variation optional. Do not add a form correction or health claim that was never spoken.

Repeated footage is not evidence of repeated narration. A later round may change “next” to “last,” refer back to an earlier demonstration or offer a different option. Review those lines in their actual round rather than pasting a previously approved caption block because the picture looks familiar.

Give each translation reviewer the final video and the block map. Check imperative wording, negation, repetitions, alternatives and whether a phrase names the current or upcoming movement. Review reading time without removing a qualification to make the caption shorter. A language track needs approval on its own timing; structural cue checks do not verify the instruction’s meaning.

Separate transition sounds from silent graphics

A beep can be the only audible signal that a passage has ended. Review meaningful non-speech audio as well as the words. W3C explains that captions include audio information needed to understand the content; DCMP offers guidance on describing relevant sounds in time with their occurrence. A human reviewer should decide what the actual sound communicates.

Do not infer a beep from a flashing timer, invent speech during music-only movement or label a sound as “rest begins” unless the recording and producer establish that meaning. Polyglot does not automatically classify transition sounds. A silent visual instruction may require a separate approved description; speech transcription does not recover it.

Check that the receiving player’s caption area leaves the movement demonstration and any essential timer visible. Inspect the narrowest supported display and larger caption text settings. Player styling varies, so a local preview is not proof of the final placement. If the destination cannot show the necessary information together, resolve the presentation with the producer before release.

Accept the exact upload and every language track

Export the format accepted by the chosen host and include the video revision and language in the filename. YouTube, for example, documents uploading a subtitle file with timing and selecting its language. It also offers its own caption creation and editing tools. Polyglot is an optional local preparation workflow, not a required intermediary or automatic upload integration.

Open the actual delivered video with each intended track selected. Check the demonstration-to-active boundary, every spoken countdown, each rest transition and the final instruction. Seek into a later repeated round, then pause and resume to verify that the selected captions still follow the player’s video clock. Read the relevant passage with sound off as a review step; do not treat this as an invitation to perform an unchecked routine.

Keep the clean master, accepted tracks and block map together. After shortening a rest, adding a demonstration or replacing narration, approve a new matching set. An old caption file with the same total duration is not necessarily aligned. If a short promotional cut is made separately, check its own words, timestamps and permissions.

Polyglot exports separate subtitle files; it does not burn text into video, author fitness-app workout data, provide live class captions or certify accessibility. This is a recorded-media production workflow. Final instruction approval and host playback acceptance remain separate checks.

Keep exploring.