Dicty
Dicty is a free, open-source dictation app under MIT. It runs whisper.cpp locally, on Metal on Apple Silicon, and the README measures median latency at 215 ms on an M3 Pro and 265 ms on an M1. The default English model, base.en, needs about 280 MB of RAM. One caveat: macOS 14 and Apple Silicon only on Mac, and Windows support is beta and needs to be built from source.
What does Dicty do?
Dicty describes itself as "Fast, private AI voice dictation for developers and professionals" with sub-300 ms latency, offline by default, and no subscription. Press a hotkey, speak, release, and the text is pasted at the cursor.
The transcription engine is whisper.cpp, Georgi Gerganov's C and C++ port of OpenAI Whisper. Dicty runs it on Metal on Apple Silicon for GPU acceleration and keeps audio and transcripts on the device. The README names base.en as the default model, with small.en as a step up. A cloud transcription path is optional.
Key facts
- Licence: MIT.
- Platforms: macOS 14 Sonoma or Sequoia on Apple Silicon (M1, M2, M3 tested); Windows 10 or 11 in beta, buildable from source.
- Speech engine: whisper.cpp, with GPU acceleration on Metal on Apple Silicon.
- Default model:
base.en, about 280 MB of RAM at runtime.small.enoption at about 620 MB. - Measured latency, from the README: median 215 ms and 95th percentile 285 ms on M3 Pro 18 GB; median 240 ms and 95th percentile 310 ms on M2 16 GB; median 265 ms and 95th percentile 340 ms on M1 8 GB.
- Minimum tested RAM: 8 GB, measured on an M1.
- Offline by default; a cloud transcription path is optional and the README says audio stays on the device by default.
- Install on macOS:
brew install --cask dicty-io/dicty/dicty, or download a disk image from GitHub Releases. - Install on Windows: build from source at the time it was read.
What does it replace, and where does it fall short?
Dicty replaces Superwhisper and Wispr Flow for Apple Silicon Mac users who want a push-to-talk hotkey, local transcription and no subscription, with published latency figures.
Where it falls short: macOS 14 and Apple Silicon only on Mac. Intel Macs and older macOS are not supported. Windows is beta and requires building from source, which is more work than most readers want. The base.en model is English only; other languages need a different Whisper model, and the README focuses on English. Cleanup is not part of Dicty; this is a transcription app.
How does Dicty run?
On macOS, install through the maintainer's Homebrew tap with brew install --cask dicty-io/dicty/dicty, or download the signed disk image from GitHub Releases. On first launch, Dicty asks for Accessibility and Microphone permissions, and the default base.en model downloads. Hold the hotkey, speak, release.
On Windows, the README describes it as beta and the install path is to clone the repository and build from source. There is no signed installer at the time it was read.
Who is Dicty for?
A developer or professional on an Apple Silicon Mac who wants fast English dictation with measured latency on their own hardware, kept on the device, under MIT. A multilingual user, a Windows user who does not want to build from source, or a Linux user should look at Plume, Handy or OpenDictate.
What limits does the README state?
- macOS 14 Sonoma or Sequoia only, on Apple Silicon.
- Windows is beta and build-from-source only.
- The default base.en model is English; other languages need a different Whisper model.
- 8 GB RAM is the floor tested; 16 GB is listed as the recommended configuration.
- The cloud transcription path is optional and off by default.
Questions people ask
Is Dicty open source?
Yes. The LICENSE file at the repository root is MIT, an OSI-approved licence. The README says Dicty is free and open source, and that audio and transcripts stay on the device by default. The underlying whisper.cpp engine is MIT, and the Whisper weights are MIT.
Does Dicty work offline?
Yes, in its default configuration. The whisper.cpp engine runs on the Mac through Metal and the README says audio stays on the device. The optional cloud transcription path sends audio to an external service and is off by default.
Can Dicty be used commercially?
Yes. MIT allows commercial use of the code without fee. The whisper.cpp engine is MIT and OpenAI's Whisper weights are MIT, so the whole on-device path is clear for commercial use. A commercial user should still check the licence of any model other than the default.
How does Dicty compare with Superwhisper?
Dicty is free, open source and focused on English dictation on Apple Silicon, with published latency numbers for each tested chip. Superwhisper is a paid Mac app with language support across many models and prompts for cleanup. Both run on the Mac. For cleanup, pair Dicty with a separate editor step or look at VoiceInk. The open-source alternatives to Superwhisper page sets Dicty beside Handy, Voz and Golosok.
How fast is Dicty in practice?
The README publishes measured latency on three Apple Silicon chips: 215 ms median and 285 ms at the 95th percentile on an M3 Pro 18 GB; 240 and 310 ms on an M2 16 GB; 265 and 340 ms on an M1 8 GB. The maintainer ran these numbers with the default base.en model on a fixed test clip, which is why the figures carry a chip name. Real-world latency depends on the clip and the system load, and the README does not claim otherwise.
Sources
- Dicty README and LICENSE: github.com/dicty-io/dicty, read 2026-11-09.
More voice dictation apps
A free, offline push-to-talk dictation app for macOS, Windows and Linux that runs Whisper or NVIDIA Parakeet on your own machine and pastes the text at your cursor, under MIT.
A GPL-3.0 macOS dictation app that runs Nemotron, Parakeet, Cohere Transcribe, Apple Speech or Whisper on the device, with command and write modes, per-app prompts, and an optional closed local enhancement model.
A dictation and meeting-notes app under MIT that runs Whisper or Parakeet locally or cloud models with your own key, detects Zoom, Teams and FaceTime calls, and labels speakers on device.
A macOS dictation app under GPL-3.0 whose README names Superwhisper and Wispr Flow as the tools it replaces, with local models, per-app modes, a personal dictionary and an AI assistant mode.
A speech-to-text app inside the Epicenter monorepo under AGPL-3.0 that records, transcribes with a provider you choose, cloud or local, optionally polishes the text, and pastes it at the cursor.
An Apache-2.0 macOS dictation app that calls Gemini 3.5 Transcribe with your own key, writes audio to disk from the first millisecond for crash safety, and inserts polished text at your cursor without an account.
Added November 9, 2026. Every claim above comes from the project's README, LICENSE or model card, read on November 9, 2026, or from the GitHub API on the date shown in the panel. Found an error? Write to reveneau@licheo.com and it is fixed in the next weekly pass. Repository: github.com/dicty-io/dicty.
Want Dicty running on your own servers?
Reveneau sets it up on your hardware, connects it to your product, and builds the software around it.