Moonshine v2: Ergodic Streaming Encoder ASR for Latency-Critical Speech Applications

Benchmark Model Rank Results
speech-recognition-on-gigaspeech-testusefulsensors/moonshine-streaming-medium#11Word Error Rate (WER): 8.14
speech-recognition-on-gigaspeech-testusefulsensors/moonshine-streaming-small#37Word Error Rate (WER): 9.08
speech-recognition-on-gigaspeech-testusefulsensors/moonshine-streaming-tiny#57Word Error Rate (WER): 12.53
speech-recognition-on-amiusefulsensors/moonshine-streaming-medium#12Word Error Rate (WER): 9.75
speech-recognition-on-amiusefulsensors/moonshine-streaming-small#18Word Error Rate (WER): 11.69
speech-recognition-on-amiusefulsensors/moonshine-streaming-tiny#47Word Error Rate (WER): 17.95
speech-recognition-on-gigaspeech-cleanedusefulsensors/moonshine-streaming-medium#13Word Error Rate (WER): 8.1
speech-recognition-on-gigaspeech-cleanedusefulsensors/moonshine-streaming-small#40Word Error Rate (WER): 9.03
speech-recognition-on-gigaspeech-cleanedusefulsensors/moonshine-streaming-tiny#54Word Error Rate (WER): 12.46
speech-recognition-on-ami-cleanedusefulsensors/moonshine-streaming-medium#14Word Error Rate (WER): 8.93
speech-recognition-on-ami-cleanedusefulsensors/moonshine-streaming-small#20Word Error Rate (WER): 10.79
speech-recognition-on-ami-cleanedusefulsensors/moonshine-streaming-tiny#53Word Error Rate (WER): 16.94
speech-recognition-on-voxpopuli-cleanedusefulsensors/moonshine-streaming-medium#19Word Error Rate (WER): 3.37
speech-recognition-on-voxpopuli-cleanedusefulsensors/moonshine-streaming-small#29Word Error Rate (WER): 3.76
speech-recognition-on-voxpopuli-cleanedusefulsensors/moonshine-streaming-tiny#65Word Error Rate (WER): 8.11
speech-recognition-on-spgispeechusefulsensors/moonshine-streaming-medium#20Word Error Rate (WER): 2.26
speech-recognition-on-spgispeechusefulsensors/moonshine-streaming-small#37Word Error Rate (WER): 2.85
speech-recognition-on-spgispeechusefulsensors/moonshine-streaming-tiny#66Word Error Rate (WER): 5.71
speech-recognition-on-earnings-22usefulsensors/moonshine-streaming-medium#29Word Error Rate (WER): 11.44
speech-recognition-on-earnings-22usefulsensors/moonshine-streaming-small#46Word Error Rate (WER): 13.09
speech-recognition-on-earnings-22usefulsensors/moonshine-streaming-tiny#64Word Error Rate (WER): 19.52
speech-recognition-on-voxpopuliusefulsensors/moonshine-streaming-medium#46Word Error Rate (WER): 8.01
speech-recognition-on-voxpopuliusefulsensors/moonshine-streaming-small#51Word Error Rate (WER): 9.26
speech-recognition-on-voxpopuliusefulsensors/moonshine-streaming-tiny#59Word Error Rate (WER): 12.97
speech-recognition-on-librispeech-test-cleanusefulsensors/moonshine-streaming-medium#52Word Error Rate (WER): 1.67
speech-recognition-on-librispeech-test-cleanusefulsensors/moonshine-streaming-small#82Word Error Rate (WER): 2.07
speech-recognition-on-librispeech-test-cleanusefulsensors/moonshine-streaming-tiny#128Word Error Rate (WER): 3.98
speech-recognition-on-librispeech-test-otherusefulsensors/moonshine-streaming-medium#85Word Error Rate (WER): 4.6
speech-recognition-on-librispeech-test-otherusefulsensors/moonshine-streaming-small#112Word Error Rate (WER): 6.37
speech-recognition-on-librispeech-test-otherusefulsensors/moonshine-streaming-tiny#133Word Error Rate (WER): 11.54