РусскийRUEnglishEN

Your first session: step by step

Overview

A session is a single listening run of a call: Whisperer listens to the audio, keeps a transcript, and answers your questions. This page shows how to start your first meeting in the overlay, what the live transcript looks like, how to ask questions in three ways, and where to find the history afterward.

Whisperer listens to two audio sources: your microphone (your voice) and system audio (the other person's voice). On macOS, system audio is captured through the "Screen Recording" permission — without it the other person won't be heard (see macOS permissions). On Windows, system audio is taken without any additional permission — you only need microphone access (see Windows permissions).

When to use it

Start a session at the beginning of any call where you need a transcript, suggestions, or translation. Before your first launch, make sure the device is paired with your account and all permissions are granted — see Quickstart.

Step-by-step

  1. Open the app. An overlay panel (GhostPanel) with a command bar (CommandBar) appears on top of your screen.
  2. Press play. Whisperer starts recording: the volume indicator (waveform) shows that audio is coming in.
  3. Watch the transcript. In the live strip (LiveTranscriptStrip), remarks are labeled by speaker: [Me] for your voice and [Them] for system audio.
  4. Ask a question — in one of three ways:
    • By text — type your question in the input field and send it.
    • From what was heard — press Enter on an empty input field: Whisperer builds a question from the recently heard conversation.
    • From a screenshot — take a screenshot so the model can analyze the image (vision).
  5. Read the answer. Answers appear in the answer area (AnswerArea) in Markdown, with page pagination 1/N and the option to expand the model's reasoning.
  6. End the session. Press pause/stop to finish — the transcript and answers are saved to history (except in no-logs mode).
  7. Open the history in the web dashboard. There you'll find the full transcript, the answers, a meeting mind map (topics, decisions, action items), and search.

Screenshots

📸 [Screenshot: overlay command bar with the play button and the volume indicator]

📸 [Screenshot: live transcript with [Me] and [Them] labels]

📸 [Screenshot: answer area with 1/N pagination and expanded reasoning]

Common mistakes

  • The other person doesn't appear in the transcript ([Them]). On macOS the most common cause is that the "Screen Recording" permission hasn't been granted — it's what captures system audio (see macOS permissions). On Windows no permission is needed for this — check that the default output device is the one the call is playing through (loopback only takes that one); see Windows permissions.
  • The "from what was heard" question didn't work. This action only triggers on Enter when the input field is empty; if there's text in the field, that text is sent instead.
  • The session wasn't saved to history. Check whether no-logs (ephemeral) mode is on: in it, the transcript and answers are deleted after the session, although the minutes are still counted.

Best practices

  • Before an important meeting, run a short test call and confirm that both labels — [Me] and [Them] — appear in the transcript.
  • Use Enter on an empty field when you need to react quickly to what was just said, without phrasing a question by hand.
  • For analyzing on-screen code, diagrams, or errors, take a screenshot — the model will factor the image into its answer.

Related articles