Motivation & Approach The idea for WhisperIt was driven by my heavy personal use of the ChatGPT voice input feature in conjunction with reading the results which to me seems to be the fastest brain-computer-interface available right now. While voice input is often faster and more natural than typing, existing voice-to-text tools are usually app-specific (e.g., in ChatGPT) or simply bad (e.g., Google STT) because good ones are too expensive to be offered for free. I wanted to build a universal, context-aware voice input assistant - something that could sit on top of any app and deliver consistent, clean, and properly formatted transcriptions.
To make this vision a reality, I am focusing on several core goals:
-
Create a floating voice input button that can be triggered anywhere, independent of the active app.
-
Build in contextual awareness, so the text is not just transcribed but formatted and styled according to the app in use (e.g., casual for messaging, formal for emails).
-
Enable real-time rewriting: Users could speak naturally, then have the system automatically reformat, clarify, or translate their input to match their intent.
-
Keep it lightweight and responsive, avoiding the latency issues typical of cloud-only transcription.
Progress & Current Status Right now, the prototype reliably:
- Transcribes spoken text accurately.
- Adapts style based on active app context.
- Allows for additional per app custom context.
- Provides a clean, user-friendly interface for controlling and editing the output.
It’s still in progress, but the core framework is working and demonstrates the potential for making voice input a truly universal tool - no matter the app, task, or language.