Add English speech-streaming INT8-wo OV IR (en/int8)

Add English speech-streaming OV IR (en/fp16)

README: add INT8 weight-only accuracy table + size/RAM/RTFx comparison

Replace int8 with weight-only quantization (accuracy-preserving; full-PTQ regressed EN WER)

Add INT8 (encoder) + FP16 variant

Add FP32 OpenVINO IR

Add FP16 OpenVINO IR

Add model card

initial commit