Private by default

Your voice never leaves your Mac.

Most dictation apps send your audio to a cloud service. Neuron Flame doesn't. Transcription runs entirely on your Mac. Telemetry is nonexistent. Optional AI cleanup is your choice — including fully-local options.

Download for Mac Read the docs

Free · Open source · GPL-3.0 · 100% on-device by default

What "on-device" actually means here

When you press your hotkey, the audio is captured, fed to either Whisper.cpp or Parakeet V2 (both running on your Mac), transcribed to text, and pasted at your cursor. The audio buffer is held in RAM for the duration of the recording and discarded after. No file is written to disk unless you explicitly save the recording from the History tab.

The transcribed text is saved to a SwiftData store at ~/Library/Application Support/com.neuronflame.app/default.store for the History feature. That database is yours — local, encrypted by FileVault if you have it on, never transmitted anywhere.

What leaves your Mac (only when you opt in)

Neuron Flame's default install talks to:

If you turn on AI Enhancement with a cloud provider, transcripts are sent to that provider over HTTPS. You control which provider. The supported list:

For maximum privacy, use Ollama. The transcript is enhanced by a local LLM and never crosses your firewall.

Telemetry

None. Zero. We don't run a server. There's no opt-out toggle in Settings because there's nothing to opt out of. Neuron Flame doesn't know how often you use it, how long your sessions are, what language you dictate in, or what model you've selected. We literally cannot tell.

Open source under GPL-3

You don't have to take our word for any of this. The full source is on GitHub — every network call, every persistence path, every permission check is searchable. We've documented the data flows in detail in the privacy doc.

Why this matters more than usual

Voice samples carry an unusually rich amount of personal context — medical conditions, financial details, names of people in your life, spoken habits. The standard cloud-dictation business model trades that context for transcription quality. Neuron Flame's bet is that local models are now good enough that you don't have to make that trade.

Dictate without surveillance.

Free. On-device by default. Open-source so you can verify.

Download Neuron Flame
FAQ

Common questions.

If everything's local, what's the cloud LLM enhancement for?

Optional polish — turning a raw transcript into a clean email, summarising rambles, formatting code. If you don't enable AI Enhancement, nothing leaves your Mac. If you enable it with Ollama as the provider, still nothing leaves your Mac. Cloud is just one of several options.

Where exactly is my data stored?

~/Library/Application Support/com.neuronflame.app/. SwiftData stores for transcripts, dictionary, prompts, and Power Mode profiles. Whisper models in a Models/ subfolder. Optional saved recordings in Recordings/. All local, all yours.

Does enabling iCloud sync send my dictionary to Apple servers?

If you opt in to iCloud sync for the custom dictionary, the dictionary entries (words and replacements) sync via Apple's CloudKit, encrypted in transit and at rest under your iCloud account. Transcripts and audio do NOT sync.

Can I audit the network activity?

Yes — run Little Snitch or LuLu and watch. With AI Enhancement off, you'll see Sparkle's update check and nothing else. With AI Enhancement on for a cloud provider, you'll see HTTPS traffic to that one provider only.