Live Captioning in Multilingual Remote Theatre Rehearsals

Live captioning in a multilingual remote theatre rehearsal is a planned communication and access layer. Speech from the rehearsal is transcribed, optionally machine-translated, and displayed on participants’ own devices or on a shared screen while work continues. It is most useful for rehearsal conversation, scene work, note-giving and documentation; it is not a substitute for a production translator, interpreter or dramaturg.

What live captioning adds to a remote rehearsal

A remote rehearsal is a networked working session, not simply a video call with theatre exercises added. When participants are in different countries, working in different first languages, or using a shared language that is nobody’s first language, captioning can make the working language of the room more visible. It allows participants to reread a note, check a term, follow a fast exchange or keep a record of decisions. For Deaf and hard-of-hearing participants, captions may be an access requirement rather than a convenience.

Terminology matters. Live captions are real-time text of speech. Translated captions add machine translation or human translation to that stream. Surtitles are usually prepared translations projected for an audience, often cued line by line. Access captions serve Deaf and hard-of-hearing participants; working captions serve rehearsal communication. One stream may not serve both purposes well, and a captioning plan should say which purpose it is addressing.

Several platforms now support live captioning and translation for theatre and events. Pockitle Live describes browser-based live captioning for theatre and spoken events without a prepared script, sharing the original language and up to ten translations depending on the plan on audience phones and venue screens; it also notes that recognition and translation errors in its demo are uncorrected. Vimeo’s live-event help describes real-time AI-translated captions in 99 languages, generated from automatic closed captions, with up to ten target languages at once and availability limited to Custom and Enterprise customers. spf.io’s theatre page describes a different emphasis: uploading a script and releasing lines as captions, switching to automatic captioning as needed, working with Zoom, StreamYard, Twitch and YouTube Live, and turning off caption history to protect copyrighted material. Tolka Live extends live speech translation into presentation tools such as ProPresenter, stage displays, OBS and audience phones, running in a browser.

Most of these products are framed around events, broadcasts or audience display. That does not make them unsuitable for rehearsal, but it does mean the production team must check what the tool does inside a private meeting, who can control it, how it stores text, and whether it supports the languages and devices the rehearsal actually uses.

Choose the source of the caption stream

Before the first call, decide where the caption text will come from. The choice affects accuracy, privacy, preparation time and the kind of rehearsal work that is possible.

A rehearsal workflow

1. Define the language and access brief. List the languages spoken in the room, the languages participants need to read, who is Deaf or hard-of-hearing, who is using a second or third language, and whether an interpreter is also engaged. Check platform limits before promising a language count. Pockitle’s up-to-ten translations are plan-dependent; Vimeo’s translated captions are limited to Custom and Enterprise customers and require automatic captioning.
2. Prepare a caption script and glossary. For scripted scenes, upload the script and release lines where the tool supports it, as spf.io describes. For devising, use automatic captioning but assign a human monitor. Add character names, unusual pronunciations, technical terms, recurring metaphors and invented words to a shared glossary.
3. Assign roles. A caption operator is not the same as a stage manager. Name someone to start and stop captions, switch languages, watch latency and flag errors. Name a language monitor, ideally a translator or dramaturg, to review machine-translated phrases that affect meaning. For high-stakes notes or access, budget for a human captioner or respeaker.
4. Test before the rehearsal. Run a short test with at least two speakers, overlapping speech, names, accents and any code-switching likely in the room. Check line breaks, phone and laptop display, shared-screen legibility and delay. Confirm whether the session is being recorded and where caption text is stored.
5. Set speaking protocols. Ask participants to speak one at a time, identify themselves when returning to a discussion, and slow down for key instructions. Caption lag changes turn-taking: do not use the caption stream as an exact timing cue for a scene. Pause after critical notes, spell uncommon terms in chat, and repeat a line when the caption visibly fails.
6. Decide how captions are displayed. Personal devices give participants choice of language and position; a shared screen gives the group a common reference. Some rehearsals need both. Pockitle describes phones and venue screens; Vimeo describes viewers choosing a language from the player’s caption menu. In rehearsal, avoid imposing one shared translation on everyone if literacy, language preference or access needs differ.
7. Protect consent, privacy and rights. Tell participants who processes the audio, whether a transcript is retained, who can access it and for how long. spf.io notes that caption history can be turned off to protect copyrighted material. Do not record a rehearsal or archive captions without consent. Check the organisation’s data policy and the platform’s retention settings.
8. Debrief and clean the record. At the end, collect errors: misspelled names, mistranslated idioms, technical terms, missed speaker changes. Save the corrected glossary and any agreed translation choices. Keep raw transcripts separate from cleaned rehearsal notes, and do not treat an uncorrected machine transcript as a definitive record.

Common problems and practical remedies

• Automatic recognition mishears accents, dialect and overlapping speech. Use a human monitor, slow down, repeat key lines and maintain a glossary. Pockitle explicitly notes that recognition and translation errors may remain uncorrected.
• Machine translation flattens register and metaphor. Do not let machine output become the production translation by default. A translator or dramaturg should review phrases that carry character, verse, humour, politeness or cultural context.
• Latency disrupts timing. Build wait time into the rehearsal. Use visual cues and spoken confirmations rather than assuming everyone has read the caption at the same moment.
• Reading competes with watching faces. Use personal devices where possible, place captions beside the gallery view, and check contrast and text size. A shared screen can dominate the room if it is the only display.
• Plan limits and unequal resources. Some translated-caption features are enterprise-only or plan-dependent. Check costs, language limits, browser support and bandwidth before the rehearsal. Have a lower-bandwidth fallback, such as captions on phones rather than a shared high-resolution display.
• Copyright and confidentiality. A rehearsal may contain unpermitted script material or developing work. Use caption-history settings, private meetings and consent procedures. spf.io’s guidance to turn off caption history is one example of a control to verify in the chosen tool.

Making captions dramaturgically useful

Captions become more useful when they are curated. For scripted work, line numbers, character names and consistent spelling help the room follow the scene. For devising, automatic captions can capture emergent language, but the raw text needs cleaning before it becomes a rehearsal note. Where the software permits formatting, use brackets for non-spoken information such as stage directions; where it does not, keep a companion document. For translation, maintain separate columns for the original text, the working translation and any agreed alternative. This makes the caption stream part of the production’s translation record rather than a disposable overlay.

The practical rule is to plan captions as you would plan a translator: assign responsibility, test the tool, obtain consent and review the output. A caption stream accurate enough for a logistics announcement may not be accurate enough for a verse scene. Match the caption source and the level of human review to the rehearsal task.

Checklist for the first multilingual remote rehearsal

• List the languages spoken and the languages participants need to read.
• Identify Deaf and hard-of-hearing participants and any interpreter provision.
• Choose automatic, script-following or human captioning for each part of the call.
• Prepare a glossary with names, pronunciations, technical terms and invented words.
• Name a caption operator and a language monitor.
• Test latency, line breaks, device display and language switching.
• Agree speaking protocols and a fallback if captions fail.
• Explain consent, recording, retention and copyright controls.
• Debrief errors and save a corrected glossary or note.