Meetings, translation & export
Once a meeting is being captured, everything happens in the minutes web app. This page walks through reading a live meeting, turning on translation, fixing or filling in individual lines, renaming a meeting, copying a transcript to your clipboard, and downloading one to keep.
The three-column view
Open the app and you get a single screen split into three columns (on a phone they collapse into one view at a time — see On your phone):
- Transcriptions (left) — every meeting you own, newest first. Click one to open it; the ⟳ button reloads the list. (Admins see everyone's.)
- Transcript (middle) — the spoken words, line by line, as they're recognised.
- Translation (right) — the translated text for each line, lined up next to the original.
The translation column stays in step with the transcript: each translated line sits on the same row as the line it came from, so you can read across.
You don't have to wait for a meeting to end. While a meeting is live, new lines stream into the transcript on their own — there's nothing to refresh.
A meeting can carry two separate audio sources — Online stream (the captured tab) and Host mic (your own microphone) — each with its own transcript and translations. When it does, a switcher in the meeting header flips both columns between them. See Two audio sources in one meeting. You can also pick which source to export, below.
Interim vs final lines
Speech recognition works in two passes:
- An interim line appears in faint italic text at the bottom of the transcript while someone is still talking. It's a best guess and it will change as more words arrive.
- When the speaker pauses, that guess is finalised into a final line — the interim text disappears and a clean, permanent line takes its place in the transcript.
Only final lines are saved, translated, and exported. Interim text is just a live preview, so don't be surprised when it shifts around before settling.
Persian and other right-to-left text
Persian (fa) is a right-to-left language. minutes detects this and flips any Persian line so it reads correctly — the text aligns to the right and the timestamp gutter mirrors to match. This applies to both the original transcript and the translation, on whichever side the Persian text appears. English (en) and German (de) read left-to-right as usual.
Timestamps
Every line carries a timestamp. In the live view, this is the absolute wall-clock time the line was spoken (for example 14:32:05), based on when the capture session joined the call — handy for matching a quote back to a moment in the meeting. Exports use a slightly different style; see Timestamps in exports below.
Translation
Translation is per-meeting and entirely optional. It runs on your own Anthropic key, which you set once under Settings → API keys. No key means no translation — the feature simply stays off, and no text is ever sent to Anthropic.
Translation (live and on-demand) always uses the meeting owner's Anthropic key. If you haven't added one yet, see Signing in & your settings to set it up. Soniox handles speech-to-text; Anthropic only ever sees text for translation.
Turn it on and pick a language
- Open the meeting.
- Click the gear icon (⚙) in the Translation column header.
- Flip Enabled on, choose an Output language — English (
en), German (de), or Persian (fa) — optionally pick a Model (Haiku / Sonnet / Opus), and click Save.
The column header updates to show what you're translating into, for example Translation → de. From then on, every new final line in this meeting is translated into that language as it arrives.
A translation toggle applies to the next capture session and to on-demand lines. Turning translation on partway through a live meeting takes effect for new lines going forward — it doesn't retroactively translate everything already on screen. Use the on-demand "translate" action below to fill in earlier lines.
You can only enable translation if an output language is chosen — there's no "translate into nothing." Pick a language first, then enable.
"Translate this line" on demand
Any line that doesn't have a translation yet shows a small translate link in the translation column. Click it and minutes translates just that one line into the meeting's output language, right away. The result is saved and stays on the line.
This is the way to translate lines from before you turned translation on, or to translate the occasional line in a meeting where you don't want full translation running.
When a translation fails
If a line can't be translated — a network hiccup, a quota limit, an Anthropic error — that row shows translation failed with a retry link. Click retry to try that single line again. Nothing else is affected, and the rest of the transcript keeps flowing normally.
If on-demand translation reports that translation is unavailable, it usually means there's no Anthropic key on the account — add one under Settings → API keys and try again.
Rename a meeting
Meetings start out named after their underlying call identifier, which isn't very readable. To give one a proper name:
- Open the meeting.
- Click the title at the top of the transcript column (it's editable — your cursor turns into a text cursor over it).
- Type a new name and press Enter to save, or Escape to cancel.
The new name shows up everywhere — the meetings list, the export filename, and any public share link. Only you (the owner) or an admin can rename a meeting.
Copy vs. export
Two ways to get a transcript out of minutes, for two different needs:
- Copy — a one-click plain-text grab, straight to your clipboard, with no timestamps. It copies the source you're currently viewing (transcript or translation, and — in a two-source meeting — whichever source the switcher is on). Perfect for pasting into an email, a chat, or your notes. The copy icons live in the Transcript and Translation column headers (on a phone, in the meeting header). See Copy the transcript or translation.
- Export — a downloadable file with options: pick the format (
txt/md/json), what to include (transcript / translation / both), timestamps on or off, and (for a two-source meeting) which source. Use it when you want a file to keep, archive, or hand off.
The rest of this section covers export.
Export
Click Export at the top of a meeting to download it as a file you can keep, archive, or hand to someone. minutes can produce three formats, and you can choose what goes in each file. (For a quick clipboard grab without timestamps, Copy is faster.)
Formats: txt, md, json
| Format | File | What you get |
|---|---|---|
txt | plain text | One line per spoken segment, with the speaker label and (optionally) a timestamp in front. The simplest, most portable form. |
md | Markdown | The same lines as txt, but with the meeting title as a heading and a small subheading naming the platform and call. Good for pasting into notes or a wiki. |
json | structured data | A machine-readable object with a meeting block (title, platform, timing, translation settings) and a segments array — every line with its text, speaker, timestamps, source language, and all translations. Use this if you want to process the transcript with another tool. |
The downloaded file is named after the meeting title (with unsafe characters stripped); a meeting that was never renamed falls back to a neutral name.
Audio source: Both, Online stream, or Host mic
If the meeting has two audio sources (see Two audio sources in one meeting), the export dialog gains an Audio source choice:
- Both — every source in one file, written as labeled sections, one per source.
- Online stream — only the captured tab's lines.
- Host mic — only your microphone's lines.
This choice only appears for meetings that actually have both sources; a single-source meeting (including any upload) exports as before, with no source picker. It composes with everything else here — pick a format, decide what content to include, and turn timestamps on or off, all independently of which source(s) you export.
Include: transcript, translation, or both
You can choose what content the file carries:
- transcript — only the original spoken text.
- translation — only the translated text (in the meeting's output language).
- both — the original line, with its translation indented directly beneath it.
In the json format every translation is always present in each segment regardless of this choice, since json is the complete structured record.
"translation" and "both" only have translated text to show for lines that were actually translated. Lines you never translated come out blank on the translation side. Translate them first (live translation or the on-demand translate link) if you want them in the export.
With or without timestamps
Exports can include a timestamp in front of each line or leave it off. In a file, the timestamp is the elapsed time since the meeting started, written as [HH:MM:SS] — so [00:03:12] is three minutes and twelve seconds in. (This is different from the live view, which shows the absolute clock time.) Turn timestamps off for a clean, prose-style transcript.
The Export button in the app gives you a one-click download of each format with transcript and translation included and timestamps on. The export endpoint itself also supports transcript-only, translation-only, timestamp-free, and per-source variants for anyone scripting downloads against the API.
Related
- Sharing a transcript — hand a read-only web page (and downloadable export) to someone without an account.
- Transcribing an audio file — drop in a recording instead of capturing a live call; the same transcript, translation, and export tools apply.