06/18/2026
I am excited to share an exciting on-device text-to-speech product a friend of mine has been building.
Mitchell Sayre, a VectorOne alum, just launched Wfloat, and I think it's one of the more interesting things to come out of the speech space this year.
Wfloat is text-to-speech that runs right on your users' devices instead of in a data center. The practical upshot: inference is free. Instead of paying per character or per request like you would with cloud TTS, you pay one flat fee per active user. Your costs stop climbing the moment your product takes off.
It's built for anything that needs a voice. AI games, assistants, tutors, agents. The models are light enough to run on almost any phone or laptop from the last decade, setup takes about five minutes, and the wfloat-tts model is already live on Hugging Face.
