Hey Product Hunt! I’m Leo, the maker of WhisprStream. I built WhisprStream because I wanted a dictation tool that matched how I actually speak. I switch between English and Chinese constantly, sometimes in the same sentence. Most dictation tools I tried either struggled with that or required sending audio to the cloud. I wanted something that could handle multilingual speech naturally while staying fast, private, and completely local on my Mac. That led me to build WhisprStream around Qwen3-ASR and MLX. Everything runs on-device, so your voice never needs to leave your Mac. While using it myself, I realized I was also repeatedly typing the same longer prompts and instructions. That became Voice Shortcuts: say a short trigger phrase and WhisprStream instantly inserts a longer piece of text you've saved. What started as a small tool for my own workflow gradually turned into something I use every day, so I decided to open-source it. I’d love to hear how you use dictation today, especially if you regularly switch between languages. Feedback, ideas, and contributions are very welcome!

whispr-stream

On-device speech-to-text for Mac, built on Qwen3-ASR via MLX

On-device speech-to-text for Mac, built on Qwen3-ASR via MLX