Underdog shows off a 32MB speech model in new demo
, updated First seen Updated Moderate coverage, tapering
Open weights
Reported by TechCrunch, Traders Union and dev.ua3 sources ↓
Underdog’s Woof model generates speech and music in a browser demo. The company says it can run on most devices without an internet connection.
Key points
- Woof has 31 million parameters and is available to download
- Underdog says Woof runs on most devices without an internet connection
- The demo lets users type text for speech and turn a tuner to generate music
Sources
- TechCrunchRead ↗Underdog shows off a 32MB speech model in new demo
- Traders UnionRead ↗Underdog unveils 32MB speech model for on-device audio generation
- dev.uaRead ↗Startup Underdog has created a local AI model for text-to-speech, which takes up 32 MB and can work without the Internet. They created a small browser game to demonstrate it.
Details
TechCrunch reports that the demo is a retro-themed game. Users enter text on a typewriter to hear it spoken, and rotate a tuner to produce different kinds of music; clicking a desk-lamp switch also triggers a sound.
Underdog says the model works in the browser and can run locally. The demo follows the company’s release of Bark, a 167 MB speech-to-text model intended for edge devices.
Written from coverage as of
Images
- TechCrunch