How to Choose Theatre Captioning Software: Access, Evidence, and Recovery

Captioning Software for Theatre: Define the Access Promise Before Choosing a Tool

In theatre, “captioning software” can mean prepared same-language captions, translated surtitles, live speech-to-text, or a combination of shared displays and personal devices. These approaches may look alike to someone seeing text on a screen, but they address different audience needs, demand different preparation and staffing, and have different limits. Start with the experience the production can truthfully offer, not the label on a product.

Captions can be part of access, but access is not a product feature that a team can simply switch on. It depends on accurate text, meaningful sound information where promised, readable presentation, audience guidance, informed choices, and people with the authority and time to correct a problem. The system should support that work; it should not conceal work that has been left unfunded or unresolved.

Describe the Audience Service in Plain Terms

• State whether each text version is translation, same-language captioning, selected sound information, language support, or another named service.
• Say which performances, seats, languages, and reading methods are covered, rather than implying a venue-wide service that has not been tested.
• Explain where people will read: a shared projection, a fixed display, a venue device, a personal browser, or a tested combination.
• Publish accurate booking and arrival information, including limitations around seating, devices, connection, brightness, and available help.
• Keep a supported non-phone option where phone ownership, connection, dexterity, vision, cost, or personal preference would otherwise exclude people.
• Let language, access, artistic, technical, front-of-house, and production leads stop an unsupported public claim within their area of responsibility.

Separate Prepared Theatre Captions from Live Speech Recognition

Prepared theatre captions follow an approved script and a live performance that may pause, repeat, skip, improvise, or change in rehearsal. Someone must make informed timing and recovery decisions. Live speech recognition turns unplanned speech into text in real time; it can fit some events, but its accuracy, speaker identification, sound context, delay, correction path, and consent questions must be tested in the actual use. A video-subtitle editor is different again: it normally prepares files for a fixed edit.

A production can use more than one method, but it should never market a prepared caption track as automatic speech transcription, or a transcription output as reviewed access writing. Name the service so audiences know what they will receive and the team knows when the promise cannot be sustained.

Protect the Text and the Decisions Behind It

• Record the source script, cuts, additions, rights, approvals, and revision status behind every caption version.
• Keep source text, translated text, same-language captions, sound descriptions, line breaks, timing order, review notes, and approvals distinguishable and traceable.
• Edit for live reading without flattening character, rhythm, humour, culturally specific meaning, or deliberate ambiguity.
• Decide how rehearsal changes reach every affected text and display method, who can approve an urgent correction, and how a reduced service is communicated.
• Maintain a usable exit path for source materials and approved versions, including retention, export, transfer, and deletion responsibilities.
• Mark passages where pauses, repetition, audience response, improvisation, or staging need human judgement rather than fixed automatic timing.

AI-assisted analysis or translation may help create an editable first draft. It does not replace caption writers, translators, access specialists, rights holders, or the production's responsibility for review and approval. A faster draft has value only when paid time exists for qualified people to correct it before an audience relies on it.

Plan Access with People, Without Extracting Unpaid Labour

Translation and accessibility may overlap, but translated dialogue does not automatically provide complete caption access. Speaker labels, meaningful sound information, reading speed, placement, audience support, and another text version may all matter. Define each public claim with appropriate language, accessibility, artistic, and production expertise.

When disabled people, Deaf or hard-of-hearing people, multilingual audience members, or other intended users contribute their time or lived expertise, explain how their contribution will be used, obtain informed consent, and keep participation voluntary. Collect only necessary data and provide fair compensation. Do not require a diagnosis or proof of disability to take part or to receive ordinary audience support. Consultation does not replace paid specialist review or rehearsal that the production should already provide.

Artistic intent and accessibility are not opposing departments. Clear choices about what belongs on a shared display, what can be selected personally, what each version contains, and what evidence does not yet support a public claim can protect both.

Test Every Reading Route in the Real Venue

• For projection or fixed displays, test sightlines, obstruction, contrast, line length, scenic competition, and the cost of moving attention between stage and text.
• For venue devices, test quantity, distribution, charging, cleaning, return, replacement, and staff help.
• For a personal browser, test entry, connection, expected battery use, privacy, language choice, screen-brightness policy, and discreet help.
• For combined methods, state which method carries each language or access promise and rehearse what remains available if one method fails.
• Check that booking pages, pre-show messages, signage, links, QR codes, and front-of-house guidance describe the same service.

No display method is inherently accessible. The evidence is whether intended audiences can find, read, and get help with the advertised service in the production's real conditions, without disclosing private information or bringing a device the venue silently assumes they own.

Rehearse Paid Operation, Handoffs, and Recovery

• Use the planned performance script, text versions, displays, audience information, connection conditions, and staffing.
• Test pauses, skipped or repeated passages, improvisation, missed advances, incorrect starting points, late edits, and loss of one display route.
• Define who may approve text, change timing, switch methods, correct public information, narrow the service, or stop it.
• Judge recovery by whether meaning and accurate audience guidance return, not only by how quickly a screen changes again.
• Prepare both primary and backup people to follow the performance, hand over, find the current place, and use the agreed alternative.
• Schedule and pay for preparation, rehearsal, standby, handoff, correction, audience support, and post-issue review, with adequate rest and cover.

A rehearsal rescued by one expert's memory, personal device, or unpaid overtime has exposed a hidden dependency; it has not proved the service repeatable. Do not make one conscientious caption operator the production's undocumented emergency plan.

Compare Systems by Evidence, Not by Label

• Speech-to-text tools may suit genuinely unscripted spoken events, where the team can support their accuracy and correction limits.
• Video-caption tools may help prepare files but may not support rehearsal revision, live timing, recovery, projection, or audience entry.
• Slide workflows can remain sufficient where text is stable, one display is well rehearsed, languages are few, and the operator is prepared.
• Installed systems can suit venues with stable infrastructure, integrated displays, trained staff, and predictable use.
• Web-based systems can suit touring, distributed review, language choice, varied venues, and a combination of projection with personal reading.

None is inherently more professional. Compare actual labour, venue dependencies, audience routes, change control, failure behaviour, data responsibility, and export or exit work. A smaller method that a team can honestly support can be safer than a broad system with no people or recovery plan behind its access promise.

Before procurement, require evidence that approved versions withstand rehearsal changes in performance conditions; every advertised reading route can be found, read, and supported; primary and backup staff can recover without an unplanned specialist; and accurate information, correction authority, data handling, and exit work are funded. If a necessary condition is missing, narrow the promise or stop procurement before audiences are asked to rely on it.

What SurtitleLive Supports, and Does Not Decide

SurtitleLive can support prepared text, language versions, collaborative review, projection, audience browser access, language selection, and human-controlled caption advancement during a performance. AI-assisted analysis and translation can help prepare drafts. It does not replace caption writers, translators, accessibility specialists, dramaturgs, or caption operators; approve scripts or final text; infer performance timing or follow actors automatically; provide staff, venue infrastructure, audience devices, power, or connectivity; certify accessibility or legal compliance; decide an artistic or access promise; or guarantee that a chosen method is reliable. The company, venue, and production retain those responsibilities and must test their chosen method in local performance conditions.

Related decisions

Compare theatre captions, subtitles, and surtitles, decide when translation and accessibility need different text, assess mobile reading alongside projection, and plan subtitle backup and recovery.