OWLS: Scaling Laws for Speech Recognition and Translation Collection 🦉 A suite of Whisper-style models from 250M to 18B parameters. Trained on up to 360K hours of data. 16k sampling rate. • 8 items • Updated May 3, 2025 • 7