01 · CONTROL
The recording stays inspectable.
You keep the source audio and can return to the exact conversation instead of trusting a cloud-generated abstraction alone.
Local by default
AI meeting notes that run on your Mac.
HiDock App uses the compute already on an Apple-silicon Mac to transcribe recordings and draft summaries. The default path does not upload raw audio to a transcription service.
What “on-device” means
A location claim you can trace.
“Private AI” can be vague. HiDock App’s default architecture is specific: the recording arrives on the Mac, the speech model reads it there, and the summary model reads the resulting text there. The core processing loop stays inside the machine.
The source recording, speaker-labeled transcript, and generated note remain available together. That makes the AI output reviewable instead of turning it into an answer detached from its source.
Your choice remains visible. If you prefer a cloud summary engine, it can remain available in Settings. Choosing that path changes the boundary, which is why it is optional rather than the default.
| Capture input | A recording transferred from a supported HiDock device. |
|---|---|
| Speech model | Runs on the Apple-silicon Mac and produces the transcript. |
| Speaker layer | Separates voices into reviewable speaker labels; labels can be corrected. |
| Summary model | Built into the app and runs locally by default. |
| Storage | Recording, transcript, and notes are kept on the Mac in the local workflow. |
Why local changes the product
The boundary shapes more than privacy copy.
02 · RESILIENCE
The core workflow is not a meeting bot.
Capture starts on HiDock hardware, so the note does not depend on adding a participant to a supported calendar call.
03 · HONESTY
Local does not mean infallible.
Smaller on-device models can trail the largest cloud systems. Check important names, numbers, dates, and commitments.
Common questions
Local AI, without magical thinking.
Does HiDock App send audio to a transcription API?
Not in the default local workflow. Transcription runs on the Mac using the recording stored there.
Do I need an OpenAI or other AI provider key?
No for the built-in local summary path. It is designed to work without a separate AI key or summary-server account.
Can I choose a cloud engine?
Yes. A cloud engine can remain available as an explicit option in Settings. Review that provider’s data terms before changing from the local default.
Why does the app require Apple silicon and substantial memory?
Speech and language models need local compute and memory. See the current requirements before downloading; 24 GB RAM is recommended.
Use the Mac you already trust.
Download the current alpha for Apple silicon.