From Zero to AI Notes: A Complete Audio Note Walkthrough

What You'll Build

Follow along, and you'll come out with Audio Note installed, the right model for your hardware, and an AI-summarized note from your first audio file. Core transcription runs on your machine; whether anything reaches the cloud depends on the AI backend you choose later.

Step 1: Download and Install

Audio Note is available for macOS and Windows. Download the latest version from audionote.app. Sign in and complete subscription authorization to start local transcription — local transcription itself does not require a cloud API key.

Step 2: Download a Transcription Model

The first time you launch Audio Note, you'll be prompted to download a model. This is a one-time cost that makes everything work offline afterwards.

Which model should you pick?

Model	Audio Note model file size	Best For
Tiny	80MB	Quick tests and older machines
Base	150MB	Lightweight tasks, short audio, temporary notes
Small	500MB	Everyday meetings, lectures, and interviews
Medium	1.5GB	Accuracy-focused work and difficult audio
Large-v3	3.1GB	Quality-first material you will archive or reuse
Large-v3 Turbo	1.6GB	Large-model behavior with stronger turnaround-time needs

This table follows the whisper.cpp / GGML model file sizes used by Audio Note. Runtime memory or GPU usage still depends on the device, acceleration path, and parameters. For a first run, start with Small or Large-v3 Turbo on modern machines; try Tiny or Base first on older hardware.

The model downloads once and stays cached. After that, transcription works offline.

Step 3: Transcribe Your First File

1. Click the File Transcription tab (or drag a file onto the window)
2. Select an audio file — supports MP3, WAV, M4A, AAC, FLAC, and video formats
3. Click Transcribe

Processing time depends on the model, hardware, audio length, and audio quality. For your first run, test a short representative clip before standardizing on a model. The progress bar shows status in real time.

When it's done, you get a full transcript with timestamps, speaker labels, and word-level highlighting.

Step 4: Generate AI Notes from Your Transcript

This is where Audio Note differs from a plain transcription tool. Instead of exporting raw text and processing it elsewhere, you can generate structured notes right in the app.

1. Open the transcript
2. Click the AI Chat button
3. Choose a preset:
   • Summarize — condenses into a few bullet points
   • Action Items — pulls out tasks, deadlines, and owners
   • Meeting Notes — structures as minutes
   • Key Takeaways — highlights the most important points
4. The AI processes through your configured backend — local models such as Ollama or LM Studio run on your machine, while cloud APIs send transcript text to that service — and results appear in the side panel

With a local AI backend, the transcript can stay on your computer; with a cloud AI backend, transcript text is sent to that service.

Step 5: Export or Continue Working

You can export the transcript and AI notes in several formats:

• Markdown — for Notion, Obsidian, etc.
• SRT / VTT — subtitle files for video editing
• TXT — plain text
• JSON — programmatic access

Or keep everything in the built-in editor. It supports rich text, tags, and search across all your notes.

Setup Checklist

Step	Check
Install	Downloaded from the official site and launches correctly
Model	Model is cached locally and can transcribe offline
First transcript	File imports successfully and timestamps appear
AI notes	You know whether the backend is local or a cloud API
Export	Format matches the destination tool

What's Next

This is the basic workflow. From here, you can explore:

• Real-time transcription — capture meetings as they happen
• Link transcription — paste a URL from YouTube or Bilibili, get a transcript
• Watch Folder — auto-transcribe new files in a directory
• System audio capture — transcribe audio output from any app on your Mac

Each of these has a dedicated guide in this blog series.