-
Notifications
You must be signed in to change notification settings - Fork 17
Description
I'm testing FireRedASR2S against some English audio, running this command:
fireredasr2s-cli --asr_model_dir ~/FireRedASR2S/examples_infer/asr_system/pretrained_models/FireRedASR2-AED --asr_use_gpu 0 --asr_use_half 0 --asr_batch_size 16 --beam_size 3 --nbest 1 --decode_max_len 0 --softmax_smoothing 1.25 --aed_length_penalty 0.6 --eos_penalty 1.0 --return_timestamp 1 --enable_vad 0 --vad_model_dir ~/FireRedASR2S/examples_infer/asr_system/pretrained_models/FireRedVAD/VAD --vad_use_gpu 0 --smooth_window_size 5 --speech_threshold 0.5 --min_speech_frame 20 --max_speech_frame 2000 --min_silence_frame 10 --merge_silence_frame 50 --extend_speech_frame 5 --vad_chunk_max_frame 30000 --enable_lid 1 --lid_model_dir ~/FireRedASR2S/examples_infer/asr_system/pretrained_models/FireRedLID --lid_use_gpu 0 --enable_punc 1 --punc_model_dir ~/FireRedASR2S/examples_infer/asr_system/pretrained_models/FireRedPunc --punc_use_gpu 0 --punc_batch_size 32 --punc_with_timestamp 1 --punc_sentence_max_length 25 --write_textgrid 1 --write_srt 1 --save_segment 1 --wav_paths {wav files}
This prints dictionaries like {'uttid': '<path to wav', 'text': 'Sentence one.Sentence two,With a comma.Sentence three.', ...}
I would expect that text to be formatted as
Sentence one. Sentence two, with a comma. Sentence three.
That is, there should be a space after each punctuation mark, and it should only capitalize the first letter of the word after .?!, not ,.