Mega-ASR

Robust automatic speech recognition for challenging real-world audio. Handles noise, far-field, echo, reverberation, and more using a foundation model trained on 2.6M samples across 54 acoustic scenarios.