Frequently asked questions

Is OpenMila affiliated with Island or with Mila?

No. OpenMila is an independent community port of Mila for Linux and Windows. It is not affiliated with or endorsed by Island Technology, Inc. Mila was created by Uri Harduf at Island and released under the Apache License 2.0; OpenMila reuses that code where the platform allows and re-implements only what is tied to macOS.

Can I use it on a Mac?

No. OpenMila does not build for macOS and does not try to. On a Mac, use the original, island-io/mila, which is the app this is a port of.

Does it need a GPU?

No. OpenMila runs on the CPU on any machine, and the machine it is developed on has no GPU. The release packages do carry whisper.cpp's Vulkan backend, so a real GPU is used where one is present, and a software-only Vulkan device is refused because it is slower than the CPU backend. None of that has been timed on real GPU hardware yet, so no speed is claimed for it. There is no CoreML or Neural Engine path off macOS.

The full requirements, minimum and comfortable, are on the download page.

Does it need the internet?

Not for transcribing. You need a connection once to download a Whisper model (1.6 GB for English and other languages, 3 GB for Hebrew), and again if you turn on speaker diarization, which fetches PyTorch for the CPU, about 200 MB. After that, recording and transcription work offline.

The AI features can be offline too. Ollama (local) is one of the built-in providers, so titles, summaries, action items and Live AI can run against a model server on your own machine at localhost, with no API key and no request leaving the computer.

A connection is otherwise used only for the update check against the GitHub releases page, and for the optional remote transcription backend and any cloud AI provider you choose. Both of those are off by default and only ever talk to the service you configure. See privacy for the full list.

What about Hebrew?

Hebrew is a first-class language here, not an afterthought. A recording is set to Hebrew or English and the model is routed accordingly: ivrit.ai large-v3 for Hebrew, OpenAI large-v3-turbo for English and other languages. Hebrew text is rendered right to left, with the direction decided per block, and there is a separate dictation hotkey for Hebrew (Ctrl+Alt+3 by default).

Is Windows ready to use?

Treat it as untested. Windows is built and tested on a windows-2025 runner in CI, where the core, the CLI, the MCP helper, the tests, the WinUI app and the portable zip all pass, but nobody has run any of it on a Windows desktop yet, so no Windows feature is marked as done. Linux is the development platform and the better supported of the two.

How do I report a bug?

Open an issue, and attach a diagnostic report from Settings > General > Export diagnostic report: system information, your settings with credentials redacted, the shape of your recordings with no titles or paths, and the log files. For questions rather than bugs, use discussions.

For a security problem, do not open a public issue: use private vulnerability reporting. For anything else, the contact form is at nx1xlab.dev/contact.

Is it free?

Yes. Free software under the Apache License 2.0, with no account, no subscription and no paid tier. If you use the optional AI features you pay whoever provides them, because OpenMila ships no API keys and has no provider of its own.