Quickstart

Install Prismical, capture transcript and scratch notes, then use skills to generate an enhanced AI note.

This guide gets the desktop app ready for real use: install Prismical, finish first-run setup, record a meeting or voice note, write scratch notes while the transcript is captured, and use a skill to generate the finished note.

Before you record

Prismical does not join meetings as a visible bot, but you are still responsible for following your local recording laws, workplace policies, and meeting consent requirements.

1. Install Prismical

Download for Mac or Windows

Use the desktop download page if you want the guided Mac architecture picker or the Windows installer.

GitHub releases

Download the latest release directly from GitHub.

Homebrew

On macOS, install with brew install --cask prismical.

Prismical is available for Apple Silicon Macs, Intel Macs, and Windows 10/11.

2. Finish First-Run Setup

Open Prismical

Launch the app and choose Get started on the welcome screen.

Grant audio permissions

Grant microphone access. On macOS, also grant system audio access so Prismical can capture meeting audio from apps like Zoom, Google Meet, Teams, Slack, WebEx, Discord, and browser calls.

Choose a transcription model

Pick the recommended local Whisper model or choose a different size. Larger models usually improve accuracy; smaller models download faster and run better on older hardware.

Let the model download

If the model is not already on disk, Prismical starts the download and shows progress. Wait for the model to be ready before relying on local transcription.

Cloud transcription

First-run setup focuses on local Whisper. After onboarding, open Settings → AI Models to connect a cloud transcription provider and make it the default for transcription.

3. Record Your First Meeting

Create or open a note

Click Note in the app header, press Cmd+N on macOS or Ctrl+N on Windows, or click Take Notes from the floating widget. If Prismical detects a meeting, the meeting alert can create a note and start recording for you.

Start recording

In the note, click the microphone button in the recording dock. Prismical records microphone and system audio together, so the transcript can separate You from Them.

Take scratch notes

Write quick bullets, decisions, names, links, or half-formed thoughts in the note while the conversation is happening. These notes become extra context for the AI pass later.

Watch the transcript

Open the transcription panel from the dock. New transcript segments appear with speaker labels and timestamps while recording is active.

Stop when done

Click the stop button when the conversation ends. The transcript remains attached to the note, and your scratch notes stay in the editor.

Transcription is useful on its own

You can stop here if all you need is a transcript. The full Prismical workflow continues by combining the transcript, your scratch notes, and a skill prompt into an enhanced note.

4. Generate The AI Note

Set a text generation model

Open Settings → AI Models and set a text generation model. Transcription models turn speech into text; text generation models power skills.

Run Enhance

Click the sparkle button in the note dock and run Enhance. The built-in Enhance skill uses your scratch notes plus the linked transcript to produce a summary-style note with useful sections such as key points, summary, and action items when the source material supports them.

Choose how to apply it

Prismical stages the skill output before changing your note. Accept it as a new appended section, use it to replace the note body, or run a different mode from the skill menu.

Customize the prompt

Open Settings → Skills to add custom skills with your own prompts. Use them for recurring formats like sales call notes, standups, research interviews, therapy notes, lecture notes, or personal journals.

5. Capture Personal Notes

You do not need to be in a meeting. Create a note, write normally, or use the floating widget's Take Notes action to start a voice note. Prismical stores the transcript with the note, so the same flow works for personal capture: rough speech or typing first, then a skill turns it into structured notes.