import gradio as gr
from asr import transcribe_multiple_files, ASR_LANGUAGES, model
from lid import identify, LID_EXAMPLES
from csv_processor import CSV_FILE_PATH
import logging
import soundfile as sf
import os
# Set up logging
logging.basicConfig(level=logging.DEBUG)
logger = logging.getLogger(__name__)
def download_csv():
file_path = CSV_FILE_PATH # Путь к вашему файлу
if os.path.exists(file_path): # Проверяем, существует ли файл
return file_path # Если файл существует, возвращаем путь для скачивания
else:
logger.error(f"file {file_path} not found!")
full_path = "/home/user/app/"+ file_path
exist_full_path = os.path.exists(full_path)
res = "found" if exist_full_path else "not found"
logger.error(f"file {exist_full_path} {res}!")
return None # Возвращаем None, если файла нет
# Prepare language options for Dropdown
language_options = [f"{k} ({v})" for k, v in ASR_LANGUAGES.items()]
bam_val = "bam (Bamanankan)"
bam_index = 0 if bam_val not in language_options else language_options.index(bam_val)
download_interface = gr.Interface(
fn=download_csv,
inputs=[],
outputs=gr.File(label="Download CSV"),
title="Download CSV file",
description="Download file audio_plus_hash_uniq_07102024.csv"
)
mms_transcribe = gr.Interface(
fn=transcribe_multiple_files,
inputs=[
gr.File(type="filepath"), # Allow multiple audio files
gr.Dropdown(
choices=language_options,
label="Language",
value=language_options[bam_index] if language_options else None,
),
gr.Textbox(label="Optional: Provide your own transcription"),
],
outputs=gr.Textbox(label="Transcriptions", lines=10),
title="Speech-to-text",
description="Transcribe multiple audio files in your desired language.",
allow_flagging="never",
)
mms_identify = gr.Interface(
fn=identify,
inputs=[gr.Audio()],
outputs=gr.Label(num_top_classes=10),
examples=LID_EXAMPLES,
title="Language Identification",
description="Identify the language of input audio.",
allow_flagging="never",
)
tabbed_interface = gr.TabbedInterface(
[mms_transcribe, mms_identify, download_interface],
["Speech-to-text", "Language Identification", "Download CSV file"],
)
with gr.Blocks() as demo:
gr.Markdown(
"MMS: Scaling Speech Technology to 1000+ languages demo. See our blog post and paper.
"
)
gr.HTML(
"""Click on the appropriate tab to explore Speech-to-text (ASR) and Language identification (LID) demos."""
)
gr.HTML(
"""You can also finetune MMS models on your data using the recipes provided here - ASR TTS"""
)
gr.HTML(
"""
for more control and no queue."""
)
tabbed_interface.render()
gr.HTML(
"""
"""
)
if __name__ == "__main__":
mms_transcribe.launch()