Functionality · 1 min read
Transcription
Record or upload audio and get text back — with speaker identification.
Transcription turns speech into text directly in a conversation. You can record live, capture computer audio at the same time, or upload a file recorded on another device.
The feature is switched on in the assistant’s AI settings. When it is enabled you will see a microphone icon in the chat field.
Record directly in Intric
Section titled “Record directly in Intric”Click the microphone icon to start a live recording. Intric captures audio from your device’s microphone.
If you also want to capture what is playing through your computer’s speakers, enable Include computer audio before you start. This is particularly useful for hybrid meetings, where both the room audio and the participants on the video call need to be picked up.
Click stop to finish. The recording is stored locally in your browser — nothing is sent to Intric until you click Send. Only then does the audio leave your device.
Upload a pre-recorded file
Section titled “Upload a pre-recorded file”If you recorded on another device, click the attachment icon in the chat field and select the audio file. It is transcribed the same way as a live recording. Which file formats work depends on the transcription model the assistant uses.
Speaker identification
Section titled “Speaker identification”If the recording contains more than one voice, the text is split by speaker, labelled Speaker 1, Speaker 2 and so on, in the order each voice is first heard. Replacing those labels with real names is not currently possible.
After transcription
Section titled “After transcription”The text lands in the conversation, and you can work with it like any other source material: ask the assistant to summarise the discussion, pull out decisions and action points, or translate it into another language.
Where the audio is processed
Section titled “Where the audio is processed”Transcription uses the transcription model configured on the assistant. The same model handles both speech-to-text and speaker identification, and the audio data stays within that model’s data-handling boundary. Which models you can choose from is governed by the Space’s security classification — so a Space intended for more sensitive material can restrict you to a model that runs within a given jurisdiction.
Read more in the help center: Transcription.
Test your knowledge
Question 1 of 2
What is required to transcribe in an assistant?