We’re building Superwhisper, a voice-to-text app for macOS, Windows, iOS, and soon Android.
Over the past year we’ve been building S1, our own family of small language models designed specifically for Superwhisper.
The models serve different parts of the voice-to-text pipeline. S1 Mini, for example, is designed for fast post-processing: taking a raw transcript and cleaning it up into the text you actually want to write.
We wrote up the S1 family, how the models work, and why we decided to build our own and contribute to OSS. Would love feedback!
We’re building Superwhisper, a voice-to-text app for macOS, Windows, iOS, and soon Android.
Over the past year we’ve been building S1, our own family of small language models designed specifically for Superwhisper.
The models serve different parts of the voice-to-text pipeline. S1 Mini, for example, is designed for fast post-processing: taking a raw transcript and cleaning it up into the text you actually want to write.
We wrote up the S1 family, how the models work, and why we decided to build our own and contribute to OSS. Would love feedback!