Underdog has released a browser-based game that showcases Woof, a compact text-to-speech model the company describes as just 31.7 MB in size. The playable demo runs entirely in the browser and uses Woof to generate all voices without routing audio to cloud servers, according to the developer.
Game demo and interaction
The game places users in a retro-themed room featuring a typewriter, a CRT monitor and a desk lamp that produces a sound when toggled. An introductory message in the game refers to “intrusive” cloud providers, and a tweet from Underdog noted the demo’s tagline and invite mechanic: “Defeat Darkcloud for an invite.” Players can type text on the virtual typewriter and hear it spoken back, and they can rotate a music tuner control to produce different musical output.
Model details and related releases
Underdog says Woof has roughly 31 million parameters and is available for anyone to download. The company also states that the model can run on most devices without an internet connection, positioning it for local assistant experiences or on-device audio tasks.
The Woof demo follows earlier releases from the same developer, including Bark, a 167 MB speech-to-text model aimed at edge transcription and dictation, and a 27-billion-parameter model that the company says can run on laptops. Underdog has also opened a waitlisted app it markets as a locally hosted alternative to Muse and Instinct.
Founder Sigil Wen framed the effort as part of a push for private, locally runnable models. Wen told reporters the goal is to train “the best open models that are American-made, that are open, that are capable and allow people to have private AI,” and predicted more capable models could run on consumer devices within months.
The demo and the model files are publicly posted by Underdog; the company’s tweet announcing the game highlighted the 31.7 MB size and invited users to try the in-browser experience.
Original source: TechCrunch AI