What changed

Underdog AI announced a new speech model named Woof, which is only 32MB in size. The model is showcased in a retro-themed game that runs entirely in the browser without requiring an internet connection. The game features interactive elements like a typewriter and music tuner, all powered by Woof. The model has 31 million parameters and is open for download.

What this means in practice

The Woof model allows users to type text on the game's typewriter, which is then spoken back by the model. It can also generate different types of music by rotating a tuner. The model is intended for local use, meaning it can run on most devices without needing to connect to the cloud. Underdog claims the model is suitable for creating local assistant experiences or personalized podcasts. The company also released a 27-billion-parameter model for laptops and a waitlisted app for local AI alternatives.

Limitations and availability

The Woof model is available for download and is designed for edge devices. It is not intended to replace larger models but may serve niche applications like local assistants or personal content creation. The supplied announcement does not specify the exact hardware requirements beyond general device compatibility. The model is part of Underdog's broader effort to develop open, American-made AI models that prioritize user data ownership.

Source facts for Underdog AI releases 32MB speech model for local use: AI companies across the world are racing to make bigger and better models, but the problem is all of them need massive amounts of processing power. On the other end, some researchers and startups are creating smaller models that can run on edge devices, like your phone or computer. One such startup, Underdog, has released a game built with its new voice model called Woof that’s only 32MB large.
Original AINewsTrend explainer based on the cited source. This graphic summarizes source statements; it is not an independent benchmark.

Sources and further reading

  1. techcrunch: Underdog shows off a 32MB speech model in new demo