AiphoriaTech/aiphoria_english_asr_v2_160ms

Model Description

This is a streaming speech recognition model for English language trained with NVIDIA NeMo.

• ASR Architecture: Hybrid Transducer-CTC FastConformer
• Input: Audio (.wav, 16kHz, mono)
• Output: Text (transcript)

ASR Metrics (Greedy, chunk_size=160ms, via calculate_streaming_metrics.py)

• Rapha EN-US Non-Accented WER - 16.83%
• Rapha EN-US Afro American WER - 18.68%
• Rapha EN-UK Non-Accented WER - 23.97%