Mime - Voice-Controlled Automation Built on Automa

[Aakash-star320]

Hi Automa team! 👋

I built Mime, a voice-controlled browser automation tool on top of Automa.

🎥 Demo:

final_mime_video.mp4

📦 Repo: https://github.com/Aakash-star320/mime-voice-automation

What it does:
Execute any Automa workflow by voice (e.g., "search Python on YouTube", "open Gmail", "fill contact form")
Records workflows with parameters once, then extracts dynamic values from voice commands
Uses OpenAI Whisper locally for privacy-first speech recognition
Example: Record "search golang on YouTube" with a parameter, then say "search rust on YouTube" - Mime extracts "rust" and runs the workflow.

Performance Note: The demo shows CPU-only performance (~30 seconds for speech recognition). On GPU-enabled systems, transcription is much faster (~2-5 seconds).

Built with Express.js, FastAPI, PostgreSQL, and Automa's workflow engine.

Thanks for creating such an extensible tool - Mime wouldn't exist without Automa's foundation! Would love to hear your thoughts. 🚀