Transcription

Transcription turns spoken audio into text that an assistant can work with. Record a meeting on the Record page, record straight from a conversation, or upload a file recorded elsewhere — then let the assistant summarize, extract actions, or translate what was said.

Recording with Record

Record in the sidebar opens a page built for meetings and dictation. You can also open it from the command menu (⌘K or Ctrl+K) by typing “record” or “transcribe”.

1. Click Start recording. Enable Include computer audio first if you also want the sound playing on your computer, such as the other side of a video call. Computer audio works in Chrome and Edge on desktop.
2. While you record, use Pause and Resume as needed, and write anything you want the assistant to know in the Notes field — decisions, names, what you want out of the meeting.
3. Click Stop recording. If you stopped too early, click Continue recording to add to the same recording.
4. Choose the assistant under Send to an assistant. The page opens on the assistant you last sent a recording to. Search by assistant or space name; pinned assistants are listed first. The space’s security classification is shown before you send.
5. Click Send. The conversation opens with the recording attached and your notes as the message, and the assistant’s own instructions decide what to do with it.

If the assistant has highlighted skills, they appear as buttons below the card. Clicking one sends the recording to that skill instead, for example a skill that writes meeting minutes.

If the page closes or the connection drops

The recording is uploaded to Intric in short pieces as you record, and each piece is kept in your browser until Intric has received it. If the tab closes, the browser crashes or you lose your connection, nothing recorded so far is lost: the recording appears under Unfinished recordings on the Record page, where you can resume or discard it. Your notes are kept with it.

How long an unsent recording is kept

A recording you never send is deleted after one day, from your browser and from Intric. Discard deletes a recording at once; Discard all clears every unfinished recording.

A recording stops automatically when it reaches the maximum file size, with a warning a few minutes before. Send what was recorded and start a new recording to carry on.

Recording from a conversation

You can still record directly in a conversation. Click the microphone icon in the conversation input to start a live recording, and click stop to end it.

Enable Include computer audio before you start to also capture whatever is playing through your speakers.

This recording is kept only in your browser until you click Send — only then does the audio leave your device and transfer to Intric for transcription. Unlike the Record page, it cannot be recovered if the tab closes first, so use Record for anything long.

Uploading a recording

To transcribe a recording made on a separate device, click the attachment icon in the conversation input and select the audio file. Intric transcribes it the same way as a live recording.

Format	Extensions
MP3	.mp3, .mp2, .mpga, .m2a, .m3a
M4A	.m4a
WAV	.wav
OGG	.oga, .ogg, .spx, .opus
WebM audio	.weba, .webm
MP4 video	.mp4
WebM video	.webm

For a video, only the sound is transcribed.

Size and number of files

Limit	Value
Size per file	Up to 200 MB
Length of a recording	No limit. Long recordings are transcribed in parts and returned as one transcript
Files per message	Up to 20, counting live recordings and any other attachments

A self-hosted installation of Intric can set a different maximum size. For the limits that apply to other file types, see Attachments.

Which model is used

Transcription uses the transcription model configured on the assistant you send the recording to. Your audio stays within that model’s data-handling boundary. If the assistant has no transcription model, the Record page warns you before you send.

The transcript

The transcript opens in a panel beside the conversation, with a player for the recording. It shows Transcription in progress… until the text is ready. From the panel you can copy the transcript or download it, along with the original audio or video. Ask the assistant follow-up questions based on the content — summarize the discussion, extract action points, or translate it into another language.

Speaker detection

When a recording contains more than one voice, Intric segments the transcript by speaker, labeling each one generically — Speaker 1, Speaker 2, and so on — in the order each voice is first heard. To put a name on a speaker, click the pencil next to their label and type the name.

Speaker detection runs alongside transcription and matches each word to a speaker. If it fails, you still get the transcript, without speaker labels.

If a transcription fails

A job that fails reports the reason rather than stopping without explanation, so you can tell an unsupported file apart from a recording that didn’t upload cleanly.

Availability

Transcription requires your organization to have a transcription model configured and available in the Space’s security classification. Record is being turned on account by account; if you don’t see it in the sidebar, use the microphone in the conversation input. If you don’t see a transcription option where you’d expect one, check with your administrator.

Transcription can also be switched off for your entire organization. Contact Intric support to have it turned off or back on; Admins can read what changes in Turning transcription off for the organization.