Set up your models
On first launch, download a speech model. Whisper Small is recommended (about 488 MB). For built-in AI summaries, also download the summary model (about 2.7 GB). You can do this later in Settings.
SUPPORT & GETTING STARTED
A clear start.
A little help along the way.
Everything you need to turn your first recording into useful notes.
Three steps to your first notes.
Record or import
Click + to record the microphone or system audio, or import an audio or video file. Grant the relevant permission when asked. Stop recording to begin transcription.
Review and take it with you
Review the Summary and Transcript tabs, play back audio, and correct the text. Export from the toolbar as Markdown, TXT, JSON, SRT, or WebVTT.
A FEW USEFUL ANSWERS
Back to
the conversation.
Where can I download PrivyVoice?
PrivyVoice is currently being tested. There is no public download yet. This website will link to the official download when it becomes available. The app requires macOS 14 or later.
Why is there no transcript?
Check Settings → Transcription for a downloaded and selected model. If you use a Remote Audio API, check its address and credentials. A failed remote request reports an error; it does not silently switch engines.
Why is my summary marked as keyword-based?
If the summary model is missing or the configured provider is unavailable, PrivyVoice creates a local keyword-based summary and explains why. Set up the model in Settings → Summaries, then regenerate the summary.
Why can’t I record system audio?
Allow PrivyVoice in macOS System Settings → Privacy & Security → Screen Recording. If recording a specific application, make sure that application is running.
Which files can I import?
PrivyVoice imports audio and video that macOS AVFoundation can read, including common WAV, M4A, and MP3 files. Video imports keep the audio only. Convert MKV, WebM, and AVI files to a supported format first.
Can I use it offline?
Yes, with the built-in engines after downloading their models. Model downloads and any remote services you configure need a network connection. Ollama and LM Studio can also run locally if set up on your Mac.
Need a hand?
Send your macOS version, PrivyVoice version, and steps to reproduce the issue. Leave out private audio, transcripts, and API keys.