Speech model v2 (CC-BY-4.0)

Speech model used by NowCaption 1.1.6 and later. This is not an app release.

nowcaption-fastconformer-en-80ms-cc-by-v2.tar.bz2: NowCaption's own fine-tune of NVIDIA's
stt_en_fastconformer_hybrid_large_streaming_multi
(transducer branch, 80 ms lookahead), as an int8 ONNX export for sherpa-onnx. Trained further on
CC-BY and CC0 English speech (People's Speech, YODAS2, AMI, VoxPopuli, LibriSpeech) with text from
NVIDIA Parakeet TDT 0.6B v2, then averaged 50/50 with the original weights. Same size, tokens and
speed as v1.

• Word errors on the ESB English test sets: 13.7% against 15.0% for v1 (meetings 21.1% against 24.5%,
  noisy speech 20.8% against 24.5%); 12.6% against 13.9% on clips never used to choose it. See
  benchmark/README.md.
• License: CC-BY-4.0, © NVIDIA; fine-tuned, exported
  to ONNX and quantized to int8 by NowCaption. Training data and changes are in the archive's README.md.
• SHA-256: 4ecb5591981f4e6bd3456d5bd1a523ae170e22465717ae59cc30c70ddf470fb8