app.py

from transformers import pipeline
import gradio as gr
import time
from video_downloader import download_video, download_video1
from moviepy.editor import AudioFileClip
from moviepy.video.io.ffmpeg_tools import ffmpeg_extract_subclip
import datetime
import os
from pydub import AudioSegment
from pydub.silence import split_on_silence

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


def process_video1(date):
    video_path = download_video1(date)

    # 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")

    # Remove the audio file
    os.remove(audio_path)

    return video_path, transcription


iface = gr.Interface(
    fn=process_video1,
    inputs=[
        gr.inputs.Textbox(label="Date with format YYYY-MM-DD"),
    ],
    outputs=[
        gr.outputs.Video(),
        gr.Textbox(lines=1000, max_lines=1000, interactive=True),
    ],
    title="Transcribe Swedish Parliament Decisions",
    desription="This app transcribes the top Swedish Parliament decision video from the given date.",
)


def process_video(date):
    # Download the video
    video_path = download_video(date)

    # Extract the first 30 seconds of the video
    short_video_path = f"short_{date}.mp4"
    ffmpeg_extract_subclip(video_path, 0, 30, targetname=short_video_path)

    # Extract audio from the short video
    audio_path = f"audio_{date}.wav"
    AudioFileClip(short_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"] + " "
        os.remove(f"chunk{i}.wav")

    # Remove the audio file
    os.remove(audio_path)

    return short_video_path, transcription


# 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),
#     ],
#     title="Romanian Transcription Test",
# )

iface.launch()