I built SayToggle for people who spend hours with ChatGPT/Claude.
The problem: Typing is slow. Voice is 2-4x faster, but most solutions require cloud or are inaccurate.
Tech stack:
• Whisper.cpp with Metal acceleration (15x real-time on M1)
• Swift + embedded Python runtime
• 100% offline processing
• Global hotkey via Carbon API (no permissions needed)
Started with accuracy and privacy goals, but seamless AI communication became the main focus. ⌥⌘C from anywhere → speak → auto-paste to ChatGPT.
$19 one-time purchase. Built for Apple Silicon Macs.
Happy to discuss the implementation or answer questions!
I built SayToggle for people who spend hours with ChatGPT/Claude.
The problem: Typing is slow. Voice is 2-4x faster, but most solutions require cloud or are inaccurate.
Tech stack: • Whisper.cpp with Metal acceleration (15x real-time on M1) • Swift + embedded Python runtime • 100% offline processing • Global hotkey via Carbon API (no permissions needed)
Started with accuracy and privacy goals, but seamless AI communication became the main focus. ⌥⌘C from anywhere → speak → auto-paste to ChatGPT.
$19 one-time purchase. Built for Apple Silicon Macs.
Happy to discuss the implementation or answer questions!