Theo Reporter

An in-person session opens on a room: the people you added standing in a loose arc on an open floor, and the record button floating at the heart of the group — one big tap to start, the same spot to stop. When someone is marked as speaking, a spotlight glides across the room to them, so a glance tells you who Theo is listening to.

Rooms — pick the set from the room pill in the header (or from Room setup): a conference room, a boardroom, a courtroom, a library, a studio, a loft, or nothing at all. It applies the moment you tap it. Every room is a photographed set, rendered once at 4K and shipped with the product — nothing streams, nothing loads late — and graded so names and captions stay readable on top of it. The image breathes so slowly you feel it more than see it, and it holds perfectly still if your device prefers reduced motion. It is atmosphere, not data — the transcript is identical in every room.

Views — from the header, keep the room and the words side by side, go full room with the transcript in a slide-in drawer, or go full transcript with the people folded down to a strip of avatars. Your pick is remembered on this device.

Faces — give each person a photo, an emoji, or a colour by tapping their avatar in Room setup. That identity follows them everywhere: the stage, the transcript, every menu.

Before you start, add the people who will be talking. Each one gets a seat on the stage and a colour, and every line they say is tagged with their name instead of “Speaker 2”.

On a video call this is automatic — Theo is a participant, so it already knows who each person is.

In the room, Theo works out who is talking from the audio itself. It is very good at telling voices apart and less good at knowing which voice belongs to which name — so while recording, tap whoever is speaking (on the stage, or on the avatar strip) and their lines are labelled as they land.

When you stop, Theo re-reads the whole recording end to end. The seats you tapped become evidence: a voice that clearly matches the lines you labelled gets that person's name. A voice Theo can't match shows up honestly as “Speaker 2” rather than a guess — use This is… on that voice (in the transcript, or in Room setup) to fold it into the right person, and every line it said moves over.

Surnames, company names, drug names, case numbers, internal jargon — anything a stranger would not spell right on the first try. Add them in setup and Theo will use your spelling instead of guessing phonetically.

This is the single biggest accuracy win available to you, and it takes about thirty seconds. A name Theo has never seen will otherwise come back slightly wrong in every single line it appears in, and fixing that afterwards is much more work than typing it once.

Click any line to edit it — while recording or after. Your version replaces what Theo heard, and it is what every export uses from then on.

Words Theo was unsure about are underlined. That is not an error flag, it is an honesty flag: it means “I would check this”. Skimming the underlines is usually enough to bring a transcript to a level you would send to someone.

You can also reassign a line — or a whole run of lines at once, from the Who said this? menu on an unattributed block — to a different speaker, which is the usual fix after an in-person recording where two people have similar voices.

A line that shouldn't be in the record — a cough, a false start, someone's aside — can be removed from the row's actions. There is an Undo right on the spot, and removing a line never touches the rest of the transcript.

Stop — the same control that started the recording stops it. That control follows the transcript, not the tab: if the browser crashed, the page was reloaded, or the recording was started on another device, the room still shows Stop and pressing it ends the session and saves everything captured so far. Any device still recording winds itself down within about a minute.

Record more — a finished transcript isn't sealed. Press the record control again and new lines append to the same transcript, same people, same word list.

Start a new recording — in the room's ⋯ menu. It opens a fresh transcript carrying everything you set up here — the people, the word list, the room, and where it's filed — so a weekly meeting never starts from a blank form.

From the same ⋯menu in the room, or from the trash that appears on any row of your transcript list. Deleting always asks twice — one tap arms it, a second confirms — and it's blocked while that room is actively recording: stop first, then delete.

If a transcript sits on Finding speakers longer than it should, the room offers Try again and Mark as ready — the second keeps the transcript exactly as it stands and files it as complete.

Plain text — speaker, timestamp, what they said. The everyday format.

Markdown — the same thing with headings and formatting, ready to paste into a doc.

Numbered transcript — fixed-width lines numbered 1–25 per page, in the shape legal transcripts are normally read in.

Subtitles (SRT / VTT) — drop straight onto a video.

Data (JSON) — every segment, every word, every timing and confidence score, for anything you want to build on top.

Turns the transcript into a real note in your workspace, so you can annotate it, ask Theo about it, and keep it beside everything else on the topic.

Reporter appears in the Theo Law sidebar, and it asks one question before you record: who is this for? Pick the client and the transcript is filed under them — so an intake call that happens before there is any matter still lands somewhere you can find it later.

A recording mode starts the session the moment you tap it, with no confirmation step, which is why the question is settled first. Until it is answered the four modes stay quiet and the page says so.

“No client” is a real answer. A partners' meeting, a dictated memo, something you just want on the record — choose No client, just record under the client picker and the recording goes ahead with nothing attached. It stays deliberately unhighlighted: the transcript is simply yours, marked Not filed in your list.

Once a client is chosen, their matters appear beneath them and you can attach one. That is the setup step for a deposition: the transcript is filed against the matter, the parties on it become the speakers, and their names — plus the caption, court, and judge — are loaded into the word list.

A client with no matters yet is not an error, and neither is a call that belongs to the client rather than to any one case — the first row in the list files it under their name alone. Either way the history below shows only the transcripts filed where you are standing, so a client with a dozen recordings stays navigable.

If you have no clients at all, the panel hands you a Create a client button that opens the same wizard as everywhere else in Theo Law. The new client is selected for you when it is done, so you can go straight back to recording.