crosswk/SayIt

What it solves

SayIt is a voice typing tool for Windows that allows users to dictate text into any application. It solves the problem of slow typing by converting speech to polished, AI-cleaned text that is inserted directly at the cursor, eliminating the need to switch windows or manually edit filler words and recognition errors.

How it works

Users trigger a shortcut to record audio, which is then processed by a speech recognition (ASR) engine. The system supports three operational modes: Local mode (using GGUF models on the user's GPU), Cloud API mode (connecting directly to providers like OpenAI or Groq), and Server mode (connecting to a self-hosted FastAPI backend). Once transcribed, the text is passed through an AI cleanup phase where a customizable prompt can remove filler words or format ideas. It can also read surrounding text to match tone and terminology.

Who it’s for

It is designed for Windows users who want a more efficient way to input text across various editors, chat apps, and browsers, as well as teams who prefer self-hosted, privacy-focused voice-to-text solutions.

Highlights

• Universal Voice Typing: Works across all Windows software via shortcuts.
• AI-Powered Cleanup: Editable prompts to remove filler words and repair errors.
• Context-Aware Writing: Ability to read surrounding text to match tone or use selected text as an editing instruction.
• Flexible Deployment: Supports local GGUF models, various cloud APIs, and self-hosted Docker deployments.
• Transparent Data Flow: Clear indicators of where audio and text are processed for privacy control.
• Local History: Searchable records of recordings with playback and re-transcription capabilities.