Spaces:
Runtime error
Runtime error
| import streamlit as st | |
| import langcodes | |
| from allosaurus.app import read_recognizer | |
| from pathlib import Path | |
| import string | |
| from itertools import combinations | |
| from collections import defaultdict | |
| def get_supported_codes(): | |
| model = read_recognizer() | |
| supported_codes = [] | |
| for combo in combinations(string.ascii_lowercase, 3): | |
| code = "".join(combo) | |
| if model.is_available(code): | |
| supported_codes.append(code) | |
| return supported_codes | |
| def get_path_to_wav_format(uploaded_file): | |
| # st.write(dir(uploaded_file)) | |
| # st.write(type(uploaded_file)) | |
| # st.write(uploaded_file) | |
| uploaded_bytes = uploaded_file.getvalue() | |
| actual_file_path = Path(uploaded_file.name) | |
| actual_file_path.write_bytes(uploaded_bytes) | |
| if ".wav" in uploaded_file.name: | |
| return Path(uploaded_file.name) | |
| if ".mp3" in uploaded_file.name: | |
| new_desired_path = actual_file_path.with_suffix(".wav") | |
| waveform, sample_rate = torchaudio.load(actual_file_path) | |
| st.info(waveform, sample_rate) | |
| torchaudio.save(new_desired_path, waveform, sample_rate) | |
| return new_desired_path | |
| def get_langcode_description(input_code): | |
| langcode = "ipa" # the default allosaurus recognizer | |
| description = "the default universal setting, not specific to any language" | |
| if not input_code: | |
| return description | |
| try: | |
| lang = langcodes.get(input_code) | |
| alpha3 = lang.to_alpha3() | |
| langcode = alpha3 | |
| description = lang.display_name() | |
| except langcodes.LanguageTagError as e: | |
| pass | |
| return description | |
| if __name__ == "__main__": | |
| # input_code = st.text_input("(optional) 2 or 3-letter ISO code for input language. 2-letter codes will be converted to 3-letter codes", max_chars=3) | |
| supported_codes = get_supported_codes() | |
| index_of_desired_default = supported_codes.index("ipa") | |
| langcode = st.selectbox("ISO code for input language. Allosaurus doesn't need this, but it can improve accuracy", | |
| options=supported_codes, | |
| index=index_of_desired_default, | |
| format_func=get_langcode_description | |
| ) | |
| model = read_recognizer() | |
| description = get_langcode_description(langcode) | |
| st.write(f"Instructing Allosaurus to recognize using language {langcode}. That is, {description}") | |
| uploaded_files = st.file_uploader("Choose a file", type=[ | |
| ".wav", | |
| # ".mp3", # TODO: convert .mp3 to .wav and save | |
| ], | |
| accept_multiple_files=True, | |
| ) | |
| for uploaded_file in uploaded_files: | |
| if uploaded_file is not None: | |
| st.audio(uploaded_file, format='audio/wav') | |
| wav_file = get_path_to_wav_format(uploaded_file) | |
| st.write(wav_file) | |
| result = model.recognize(wav_file, langcode) | |
| st.write(result) | |