ads

New best story on Hacker News: Show HN: Kitten TTS – 25MB CPU-Only, Open-Source TTS Model

Show HN: Kitten TTS – 25MB CPU-Only, Open-Source TTS Model
818 by divamgupta | 328 comments on Hacker News.
Kitten TTS is an open-source series of tiny and expressive text-to-speech models for on-device applications. We are excited to launch a preview of our smallest model, which is less than 25 MB. This model has 15M parameters. This release supports English text-to-speech applications in eight voices: four male and four female. The model is quantized to int8 + fp16, and it uses onnx for runtime. The model is designed to run literally anywhere eg. raspberry pi, low-end smartphones, wearables, browsers etc. No GPU required! We're releasing this to give early users a sense of the latency and voices that will be available in our next release (hopefully next week). We'd love your feedback! Just FYI, this model is an early checkpoint trained on less than 10% of our total data. We started working on this because existing expressive OSS models require big GPUs to run them on-device and the cloud alternatives are too expensive for high frequency use. We think there's a need for frontier open-source models that are tiny enough to run on edge devices!

New best story on Hacker News: If you're remote, ramble

If you're remote, ramble
954 by lawgimenez | 452 comments on Hacker News.


New best story on Hacker News: Show HN: Draw a fish and watch it swim with the others

Show HN: Draw a fish and watch it swim with the others
864 by hallak | 220 comments on Hacker News.
Made this website as an exercise in vibe-coding and GCP. It was posted about a few times around the internet, on sites like [Morning Brew]( https://ift.tt/3tBerca ), [MetaFilter]( https://ift.tt/Z7vMTeK ), boingboing.net, etc. I think it's cute! I built a basic CNN trained against penises and swastikas, and then anything that doesn't hit the 63% confidence score gets sent to a mod queue, a [vibe-coded fish-tinder]( https://ift.tt/Bs8Mjbz... ). Was a fun exercise, spent about a month on it. Frontend is HTML5 hosted on github pages, backend is Node.JS on GCP.