Branching video subtitles. Check every path.

Prepare recorded clips for a branching lesson

A branching lesson lets a learner choose what happens next. Its videos may share an introduction, lead to different responses and return to the same scene. A caption track that works for one complete demonstration will not necessarily fit these separate clips.

Polyglot can transcribe authorized recordings locally on an Apple Silicon Mac, let you review the words with playback, optionally translate them and export SRT, WebVTT or text. Use one project per distinct clip and revision. Build the branching lesson in your authoring system; Polyglot does not create decision trees, attach tracks to course nodes or track learner progress.

H5P’s Branching Scenario documentation describes choices between content nodes and supports Video and Interactive Video content. Its Video library defines WebVTT track files with a label, language and track kind. This is a documented receiving format, not proof that every installed H5P version or embedded video provider handles captions identically. Confirm the controls available in your actual course before preparing the files.

Make a node-to-file map before transcribing

Give each video node a stable identifier outside Polyglot. Record its exact media filename, revision, duration, spoken language and approved caption files. Keep the learner’s choice label and the destination node alongside that entry. Labels such as “Try again” can occur in several places; a filename based only on that label is easy to attach to the wrong scene.

Two nodes may use the same unchanged clip. They can share the same reviewed caption file if the media and language really match, but test both entry routes. Two clips with similar words or the same duration are not automatically interchangeable.

This fictional equipment-return lesson shows the distinction. Times in each caption file start from that file’s associated clip, not the accumulated time a learner has spent in the course.

Fictional node map (r3)

Intro
  return-intro-r3.mp4
  return-intro-r3.en.vtt
Ask for a receipt -> Ask
Check the item -> Check

Ask
  return-ask-r3.mp4
  return-ask-r3.en.vtt
Check
  return-check-r2.mp4
  return-check-r2.en.vtt

Ask cue: 00:04 in Ask.
Do not add Intro's duration.
Replay: recheck the first cue.

Keep captions on the clip’s own timeline

Import each final video clip and review its source-language draft against that recording. Do not concatenate the captions into one course-length timeline: different choices produce different sequences, and a replay may revisit the same clip. A cue heard four seconds into a branch belongs at four seconds in that branch’s track.

If a clip was cut from a longer approved recording, prepare its track against the extracted clip or use the separate clip-trimming workflow to transform an existing track. Check cues crossing the cut and confirm the opening and ending aloud. Merely renaming the original full-length subtitle file will not adjust its timing.

Review negations, alternatives and conditional wording carefully. “We can accept it if the seal is intact” should not become an unconditional promise. Preserve the performed dialogue even when it depicts an intentionally poor response; explain the lesson’s evaluation in separate feedback. Ask the training owner to resolve a recording error instead of silently rewriting what the character said.

Review choice text separately from spoken dialogue

Questions, answer buttons, scores and feedback may be authored as interface text rather than spoken in a clip. Maintain and translate those strings in the course authoring system. Polyglot’s subtitle translation does not translate the surrounding lesson or keep its branching logic aligned.

Use a shared terminology note so a translated choice leads to the response the learner expects. Have a competent reviewer check both the track and the surrounding choice text. A subtitle saying “request a replacement” beside a button saying “issue a refund” can make the activity misleading even when each file passes a format check.

Check meaningful non-speech audio and any needed speaker context manually. W3C’s caption guidance covers more than spoken words. If extra caption events or positioning are needed, finish them with an appropriate caption editor. Subtitles alone do not assess the accessibility of branching controls, feedback, keyboard navigation or the whole course.

Attach the right track and test every reachable transition

Export the format your receiver requires. For an H5P Video element using its documented track-file field, that is WebVTT. Associate the file with the matching video, choose an accurate label and language, and use the intended caption or subtitle kind. An SRT renamed with a .vtt extension is not a conversion. Embedded YouTube or other provider videos may require the provider’s own caption workflow.

In the published learner preview, turn captions on and follow every reachable transition. Check the first and final cue of each clip, a cue in the middle, pause and resume, replay, and any route back to an earlier node. Confirm that the old clip’s text clears when a new node opens. Test each language and check whether caption selection persists across transitions; do not assume it does.

Keep a small acceptance sheet with the source node, choice, destination, media revision, selected language and result. A successful direct preview of a clip is useful, but it does not verify the route that opens it. Review a narrow viewport and keyboard access in the actual host as well.

When one clip changes, update every node that refers to it and review the affected source and translated tracks. Recheck its incoming and outgoing transitions, including loops and endings. Keep approved old versions identifiable until replacement is verified. A successful local export is not evidence that the learning system is serving the new file.

Clear the recording and its intended audience

Use recordings you own or have permission to process, translate and distribute. Confirm performer and training-owner approval, any institutional AI rules, and rights in included screens, music and third-party material. Remove confidential material through the approved media-editing process before creating a shareable version; deleting a caption does not redact the recording.

Polyglot processes media and transcript contents locally after model setup. Hosting the lesson is a separate disclosure and access decision. Follow the organization’s approved upload and retention process, and verify learner access without widening the audience. The course owner remains responsible for instructional accuracy, permissions and final accessibility review.

Keep exploring.