From Zero to AI Notes: A Complete Audio Note Walkthrough
Install Audio Note, pick a model for your computer, transcribe your first file, and turn it into AI notes. Covers model families, the free plan, and privacy.
What this walkthrough gets you
By the end you will have installed Audio Note, downloaded a model that suits your computer, turned an audio file into a transcript, and had the AI assistant turn it into notes you can export.
Core transcription runs on your computer. Whether anything goes to the cloud depends only on the backend you pick for the AI assistant. Expect about ten minutes, mostly model download time.
Step 1: Download and install
Audio Note runs on macOS and Windows. Get the latest version from the download page, install it, and sign in. Local transcription needs no cloud API key.
The free plan is enough for the file transcription in this guide. Live transcription, GPU acceleration, and more models need a paid plan; see the pricing page.
Step 2: Download a transcription model
Download one model before your first transcription. It stays on your computer, and transcription after that does not need the internet.
Audio Note has three model families:
| Family | What it is | Best for |
|---|---|---|
| Whisper | GGML models on whisper.cpp, with GPU acceleration | Files, links, watch folders |
| Realtime models | Low-latency Sherpa-ONNX models | Live microphone and meeting transcription |
| ML engine (advanced ASR) | Offline models such as Qwen3-ASR, SenseVoice, Parakeet | File transcription where accuracy or a specific language matters |
Whisper is enough to start. Sizes in the app:
| Model | File size | Good for |
|---|---|---|
| Tiny | 80MB | A quick try, older computers |
| Base | 150MB | Short audio, quick notes |
| Small | 500MB | A starting point for meetings, classes, interviews |
| Medium | 1.5GB | Accuracy matters, harder audio |
| Large-v3 | 3.1GB | Quality first, material you keep |
| Large-v3 Turbo | 1.6GB | Large-model results, faster |
File size is not runtime memory, and speed depends on your computer and acceleration. Start with Small; on an older computer, try Tiny or Base first. The free plan includes Tiny, Base, and Small (multilingual and English-only versions). On Standard or above, mostly Chinese audio is also worth trying with SenseVoice in the ML engine. Model advice: model selection tips.
Not sure? Let the app test your computer
The home screen has a Performance Benchmark. It runs a roughly 35-second English sample through Whisper Base and a lightweight realtime model, records speed, error rate, and CPU and GPU load, and combines that with your hardware (chip, GPU, VRAM, memory):
- Acceleration: Metal on Apple Silicon, CUDA on NVIDIA, Vulkan on discrete AMD GPUs; CPU on Intel Macs, Windows machines with only integrated graphics, and virtual machines. If the GPU turns out slow in the test, it falls back to CPU.
- Whisper model for files: it picks from Tiny, Base, Medium, and Large-v3 only. Medium or Large-v3 is recommended only when both tests look strong on a GPU (more VRAM, memory, or a newer chip leans toward Large-v3). Most machines get Base; machines that struggle get Tiny.
- Whisper or a lightweight model: if the lightweight model is clearly faster on your machine with a similar error rate, it becomes the file recommendation. For live transcription it works the other way: Whisper is recommended only when it is fast, accurate, and leaves CPU headroom; otherwise a lightweight realtime model.
Click Apply Recommended Settings to use the result. Model lists then mark the recommended model and flag Whisper models heavier than your machine handles well. If the recommended model is not in your plan, pick the closest one that is. Treat it as a starting point and test with your own audio.
Step 3: Transcribe your first file
- Click "Transcribe File" on the home screen, or drag a file into the window.
- Pick a file. Common audio such as MP3, WAV, and M4A, and video such as MP4, all work.
- Check the model and language, then start.
Speed depends on the model, computer, and audio length; a progress bar shows where it is. Try a one- or two-minute sample first before long files.
You get a timestamped transcript; click text to jump to that point in the audio.
Step 4: Let AI turn it into notes
This is where Audio Note differs from a plain transcriber: no copying text into another tool.
- Open the transcript.
- Go to the AI assistant.
- Pick a preset prompt such as summary, meeting minutes, or action items, or write your own.
- When it finishes, write the parts you want back into the note.
The AI assistant uses the backend you configured:
- Local (Ollama, LM Studio): the model runs on your computer and text stays there.
- Cloud (OpenAI, Anthropic, Gemini, DeepSeek, and others): transcript text is sent to that service.
For sensitive material, check you are on a local backend. More in AI summaries and action items.
Step 5: Export or keep editing
Paid plans export 11 formats:
- Markdown for Notion or Obsidian
- SRT, VTT, ASS, SSA, SUB, SBV, SMI, LRC for subtitles
- Plain text
- JSON for scripts
The free plan currently exports plain text only. See the subtitle formats guide.
Checklist
| Step | Check |
|---|---|
| Install | Downloaded from the official site, launches and signs in |
| Model | Downloaded locally; transcribes with the network off |
| First transcript | File imports; timestamps show |
| AI notes | You know whether the backend is local or cloud |
| Export | The format fits the destination |
FAQ
Does transcription need the internet? Not after the model is downloaded. Sign-in, license checks, and cloud AI features do go online.
Transcription is slow. Use a smaller model, or enable GPU acceleration on a paid plan. NVIDIA GPUs on Windows can use CUDA.
Chinese results are poor. Set the language to Chinese. Try Medium or Large-v3 Turbo, or models that do well on Chinese such as SenseVoice and Qwen3-ASR.
Can I batch files? Yes, from the batch transcription entry on the home screen. For continuous processing, use watch folders.
Free vs paid? Free: Whisper Tiny to Small on CPU, 1 GB per file, plain-text export, one watch folder. Standard adds live transcription, GPU acceleration, all official Whisper models, and all export formats. For the ML engine, Free has none, Standard gets SenseVoice and Moonshine, and Pro gets every ML engine model. Pro also adds community models, vocal separation, and ambient denoise. See the pricing page.
What to try next
- Live transcription for meetings and classes: local meeting transcription
- Link transcription for YouTube or Bilibili: video link transcription
- Watch folders: automate transcription
- Global realtime (beta) for anything your computer plays
