What shipped, and when

Translation on the venue screen

• Translated captions land about 1.3 seconds after the word is spoken (2.5 s in the slowest one in ten). The worst-case wait before a sentence was even sent for translation dropped from 6.9 s to 3.7 s.
• Every translated line stays on screen for at least 1 second, longer for longer lines, and is written continuously — no pause between lines.
• Rows are capped at 44 characters, two balanced rows per sentence, never a lone word on the bottom row.
• Nothing is lost: a translation that arrives late is still shown, and the previous sentence no longer flashes back after a silence.
• No invented full stops on sentences that were cut mid-way.

The console

• Audio is a list of sources — microphone or desk, and the computer’s own sound — each with its device or desk channel in the row. Simple and Advanced modes are gone.
• An audio interface with several inputs is picked from the row, input by input.
• Adding a source tells you what will happen before you press it.
• Every room link says who it is for — audience, organiser, moderator — and one button copies a ready-to-send message with all three.
• A window shows what a new version brings before you restart — never while on air.
• The right-hand column is now Console; Clear is a double press, no system dialog.

Style and languages

• Adjustable line spacing in the Style panel.
• Non-Latin scripts get the right typeface first in the font picker — Arabic, Hindi, Thai, Japanese and others.
• Simplified and Traditional Chinese are now two languages, on stage and on phones.

Also

• AI minutes keep 29 of 30 stated facts on a seven-hour event, up from 18.
• Scaption Access on iPhone opens with its buttons in view.

Five changes, all of them reported from a live desk rather than from a test.

• The caption window is a fixed two rows again. It used to grow and shrink between one and four rows as sentences arrived, which made the whole block heave up and down on the venue screen. It now fills, rolls up, and starts the next line underneath.
• The last sentence leaves the screen. After five seconds of silence the caption clears instead of hanging there through the applause and the next speaker walking on.
• Translations can no longer skip a line. A burst of finished sentences used to race each other; they are now paced so every line is shown in order.
• Scaption Access no longer waits for other people’s languages. A sentence still being spoken now reaches phones reading the floor language immediately, instead of waiting for a translation that reader never needed.
• Signing in is faster when opening Access from a QR code.

An audio input can now listen to a microphone, to your computer’s own audio, or to both at once. That one field is the whole of online-event support.

• Every platform works the same way, including whichever one appears next. SCAPTION listens to the sound your machine is already playing, so there is no per-platform integration and nothing for anyone to approve.
• No bot joins the meeting. Nothing to admit, nothing to block, and no second participant tile in the recording.
• Hybrid events run on any plan. The both setting sums the room microphone and the call into a single channel, so a hybrid event does not require a multi-microphone plan — the free trial can run one.
• Attendees on the call get the same Scaption Access link as the people in the room: fifty simultaneous languages, each person choosing their own. Every meeting platform’s built-in captions render exactly one.

The audience already has SCAPTION open on their phones to read captions. Now they can use it to take part.

• Ask a question from the phone. Other attendees upvote; a moderator picks one and puts it on the venue screen for whoever is on stage to answer out loud — which the caption pipeline then transcribes and translates like any other spoken line.
• Live polls — yes/no, yes/no/abstain, or A/B/C/D — with the tally drawn on screen as votes arrive, and an optional countdown.
• The moderator surface is a web page built for an iPad, not the desktop app. A moderator can run it from a signed link without having an account at all.
• Questions and poll results are folded into the AI summary, so the minutes record what the room actually asked and how it voted.
• Both the question card and the poll card can be placed anywhere on the same nine-point grid as the captions and the QR panel.

Put a QR code on the venue screen, or print the room code on the programme. Anyone in the room opens it in a browser and follows the event in the language they choose.

• No app, no account, no sign-in. A browser and the room code.
• Fifty languages at once. Picking a language nobody else picked starts translating it for that one person.
• Never a surcharge — not per attendee, not per language. That is a deliberate commitment, not an introductory offer.
• Optional read-aloud, spoken by the phone itself through headphones. Free and uncapped.
• Nothing is stored. The live captioning path writes no transcript anywhere.

The room code is the only credential, so anyone holding it can read along — which is exactly right for an open audience, and why a confidential session should use the venue screen alone.

One button turns a captioned event into a professional set of minutes: an up-front overview, topics, decisions, an action-item table with owners and due dates, figures and commitments, open questions, audience votes, and verbatim quotes — written in whichever of the fifty languages you ask for, headings included.

• Available from the desktop app next to Save, and from the Sessions list in your dashboard.
• Included in every paid plan, with no metering and no per-summary charge.
• Long events are handled properly: an eight-hour conference is summarised in full rather than truncated.
• Real wall-clock times are cited where they earn their place, and on multi-microphone events the speaker names you typed are used instead of “Input 2”.
• This is the only place transcript text is stored at all, and only when you ask for a summary. It is deleted automatically after 30 days.