app.py

from transformers import pipeline
import gradio as gr
import time
from video_downloader import download_video
from moviepy.editor import AudioFileClip
import datetime
import os
from pydub import AudioSegment
from pydub.silence import split_on_silence

pipe = pipeline("automatic-speech-recognition", model="Artanis1551/whisper_romanian3")


def process_video(date, live):
    # Parse the date to the format yyyymmdd
    date = datetime.datetime.strptime(date, "%Y-%m-%d").strftime("%Y%m%d")

    # Download the video
    video_path = download_video(date)

    # Send the video path to the live output
    live.update(video_path)

    # Extract audio from the video
    audio_path = f"audio_{date}.wav"
    AudioFileClip(video_path).write_audiofile(audio_path)

    # Split the audio into chunks
    audio = AudioSegment.from_wav(audio_path)
    chunks = split_on_silence(audio, min_silence_len=500, silence_thresh=-40)

    # Transcribe each chunk
    transcription = ""
    for i, chunk in enumerate(chunks):
        chunk.export(f"chunk{i}.wav", format="wav")
        with open(f"chunk{i}.wav", "rb") as audio_file:
            audio = audio_file.read()
        transcription += pipe(audio)["text"] + "\n "
        os.remove(f"chunk{i}.wav")

        # Update the live output with the current transcription
        live.update(video_path, transcription)

    # Remove the audio file
    os.remove(audio_path)


iface = gr.Interface(
    fn=process_video,
    inputs=gr.inputs.Textbox(label="Date with format YYYYMMDD"),
    outputs=[
        gr.outputs.Video(),
        gr.Textbox(lines=1000, max_lines=1000, interactive=True),
    ],
    live=True,
    title="Romanian Transcription Test",
)

iface.launch()