PlatformNSFW
Hedy runs full meeting pipeline offline on Apple Silicon with Qwen 3.5/3.6
Hedy's meeting app now runs its full pipeline—speech recognition, summaries, notes, chat, and coaching—entirely offline using llama.cpp and Qwen models up to 35B, with Metal acceleration and BYOM support for any GGUF weight.
Hedy, an AI meeting app, now runs its entire pipeline—speech recognition, summaries, detailed notes, chat, and live coaching—offline using llama.cpp and Qwen models. A demo this week showed the app generating meeting summaries on an M4 Max with Wi-Fi disabled. Speech recognition already ran on-device via whisper.cpp and Parakeet; the new release extends that to the rest of the stack.