DevEmperor/DictateKeyboard

What it solves

Dictate Keyboard replaces traditional typing with high-accuracy AI-powered dictation and transcription. It allows users to input text into any Android app using a variety of speech-to-text engines, reducing the friction of manual typing and providing tools to refine text through AI rewording.

How it works

Built as a standalone keyboard on top of FlorisBoard, the app integrates multiple speech engines (such as OpenAI's Whisper, Google Gemini, and Deepgram) and LLMs for text processing. Users can choose how to power the system: using their own API keys, a self-hosted server, on-device models for offline use, or prepaid credits via Dictate Cloud. It functions as a system-wide voice input, meaning it can be used even when other keyboards are active via a floating dictation button.

Who it’s for

Android users who prefer speaking over typing, individuals requiring high-accuracy transcription in multiple languages, and privacy-conscious users who want control over where their audio data is processed (on-device or via specific providers).

Highlights

• Flexible Engine Choice: Supports a wide range of providers including Whisper, Gemini, Deepgram, and Soniox, with options for real-time streaming transcription.
• On-Device Processing: Offers downloadable models (like Whisper and NVIDIA Parakeet) for completely offline, private dictation.
• AI Rewording: Integrated LLMs from OpenAI, Anthropic, and Google Gemini to translate, summarize, or change the tone of selected text.
• Full Keyboard Functionality: Includes glide typing, next-word prediction, and extensive layout customization beyond just voice input.
• Privacy-First: No tracking and allows users to bring their own API keys or use self-hosted servers to avoid third-party data storage.
• Accessibility: Specifically designed for screen reader (TalkBack) compatibility and high-contrast themes for visibility.