Press your hotkey
Start dictation from the floating desktop panel or a global shortcut.
Local-first AI dictation
Speak naturally. SpeakMagic writes privately.
Dictate into any desktop app, clean up rough thoughts, capture meetings, and search your notes with local speech and AI models running on your machine.
macOS
Windows
Linux
Offline transcription
Listening locally
Transcript
Action items from today's planning call
Follow up with design on onboarding, prepare Linux test notes, and turn the launch checklist into a customer-facing download page.
Local modelWhisper base
Paste targetAny app
SearchSemantic notes
Built for private work
Your voice stays in your studio.
SpeakMagic is designed around local transcription first. Download a model once, press your hotkey, and turn speech into text without sending every thought to a hosted API.
- Offline speech-to-text with Whisper.cpp and NVIDIA Parakeet support
- Local notes search powered by MiniLM and Qdrant
- Bring cloud features back later only when your product is ready
Daily workflow
From voice to polished text in three steps.
Speak like you think
SpeakMagic captures your voice and transcribes it locally with your selected model.
Paste, save, or summarize
Send clean text to any app, turn calls into notes, or search your private archive.
What ships in the desktop app
A complete private dictation workspace.
Universal dictation
Dictate into browsers, editors, CRMs, docs, chat apps, and anywhere paste works.
Local model picker
Choose Whisper.cpp or Parakeet models depending on speed, accuracy, and language needs.
Meeting notes
Record calls, capture speaker-aware transcripts, and keep searchable notes in one place.
Semantic search
Find notes by meaning, not just keywords, with a local vector index.
Custom vocabulary
Add names, product terms, and niche words so transcription fits your work.
Cross-platform packaging
Installer targets for macOS, Windows, and major Linux desktop formats.
Desktop native
One product across the machines your customers use.
SpeakMagic packages native helpers for hotkeys, clipboard paste, audio capture, local inference, and search so the app feels like part of the operating system.
macOSApple Silicon and Intel builds
WindowsInstaller and portable targets
LinuxAppImage, DEB, RPM, and tar.gz
First public line
Download SpeakMagic v0.1.0.
Releases are published on GitHub. Choose the installer for your platform and start with local models for private AI dictation.
Questions
Good things to know before launch.
Does SpeakMagic work without cloud services?
Yes. The product is currently positioned around local-first dictation and private notes.
Where do customers download it?
From GitHub Releases after the repository or release assets are public.
Can cloud features return later?
Yes. The desktop app keeps cloud code gated behind a feature flag for a future launch.