Interfaze open-sources diffusion-gemma-asr-small ASR model

The diffusion ASR model transcribes audio in six languages using a parallel denoising decoder instead of autoregressive generation. It is described as the first multilingual audio diffusion ASR model.