Privacy Policy

Effective July 22, 2026

The short version: Speech-to-Markdown collects nothing. There is no account, no analytics, no server we operate, and no data sent to us — because the app has no backend to send it to. Speech-to-text and document formatting run entirely on your Mac or your iPhone/iPad. No third-party AI service is used on iPhone or iPad — all AI processing there is Apple Intelligence, running on-device.

What this app does with your voice and text

Audio you record is transcribed and your documents are formatted on-device:

macOS: transcription runs locally via whisper.cpp; formatting is handled by an OpenAI-compatible LLM server that you install and run yourself on your own Mac (e.g. omlx, Ollama, LM Studio, llama.cpp). By default this server listens on 127.0.0.1 — your own machine — so nothing leaves the device. If you deliberately point the app at a remote server you control, your transcript is sent only to that address; we have no visibility into it and never see it.
iOS / iPadOS: transcription uses Apple's on-device Speech framework (SpeechAnalyzer) and document formatting uses Apple Intelligence (on-device Foundation Models). No audio, transcript, or document text is sent to Apple, to us, or to any third party — the entire pipeline runs in the app's own process on your device. The app works fully offline, including in airplane mode.

Third-party AI services

Speech-to-Markdown does not use, and does not send your data to, any third-party AI service — not OpenAI, not Anthropic, not Google, not any cloud vendor.

On iPhone and iPad, every AI feature is Apple Intelligence, running on-device. There is no server-side component, no API key, and no network request involved in transcription or formatting.
On Mac, the local LLM server is not a third-party service either: it's software you choose, install, and run yourself on your own computer. "OpenAI-compatible" describes only the API request/response format that server speaks — the same format many local-LLM tools implement so they're interchangeable — it is not a connection to OpenAI, the company. Nothing is ever sent to OpenAI.

What we collect

Nothing. Speech-to-Markdown has no analytics SDKs, no crash reporters, no advertising identifiers, no third-party trackers, and no user accounts or sign-in.

Where your data lives

Recordings, transcripts, and documents are stored only in local files on your own device — under ~/.stmd on macOS, and in the app's sandboxed container (visible in the Files app) on iOS. Nothing is synced to any server we run.

Permissions the app requests

Microphone — required to record what you say.
Accessibility (macOS only) — required so Global Dictation can type the transcribed text at your cursor in whatever app has focus.
Speech Recognition (iOS only) — required to use Apple's on-device SpeechAnalyzer.

None of these permissions are used to collect or transmit data anywhere; they exist solely to let the app function locally.

Children's privacy

Speech-to-Markdown does not knowingly collect information from anyone, including children under 13 — it does not collect information from anyone at all.

Changes to this policy

If this policy ever changes, the updated version will be posted on this page with a new effective date.

Contact

Questions about this policy: open an issue on GitHub or email xajik0@gmail.com.