Speech to Text

Turn your voice and audio recordings into text, right on your computer.

Made for Fedora with the GNOME desktop (tested on Fedora 44).

What you can do

• Dictate using a keyboard shortcut and a small floating panel.
• Transcribe audio files, detect the language, or translate to English with Whisper.
• Search your transcription history and export your text.
• Optionally polish text with an AI service you choose.
• Choose your interface language and a light, dark or system theme.

Whisper is the default engine; Parakeet and Qwen3-ASR are optional local engines. Download a model once, then transcribe offline. Vulkan GPU acceleration is included, with CPU fallback. For repeated dictation, Parakeet keeps its model in memory. Switching engines, cancelling an active transcription or quitting releases it; Whisper also reuses its voice detector.

Install

Download the Fedora x86_64 RPM from GitHub Releases, then run:

sudo dnf install ./speech-to-text-1.0.0-*.x86_64.rpm

Open Speech to Text from GNOME's app launcher. Choose a model in Recognition → Model, then record or open an audio file. Configure dictation shortcuts and paste permissions in Dictation & Audio → Dictation. The tray icon is optional; GNOME needs an AppIndicator extension to display it.

Privacy

Audio is processed locally. Settings and history stay on your computer, and API keys are stored in the system keyring. Optional cloud AI sends transcript text to your chosen provider. The startup update check contacts GitHub and can be disabled in Settings. See PRIVACY.md.

Build and help

For a local Fedora RPM, run bash scripts/build-rpm.sh; the result is always saved in dist/rpm/. Build requirements and developer checks are in CONTRIBUTING.md.

What's new · Security · Website · Contact

See LICENSE and third-party notices.