app.py

import base64
import json
import os

import modal
import requests
import streamlit as st
from loguru import logger
from pydub import AudioSegment

# password = os.environ["PASSWORD"]
run_transcription = modal.lookup("ffpub-transcription", "run_transcription")

st.set_page_config(page_title="Speech to Text Transcription App")


@st.cache
def transcribe(url, audio_b64):
    return run_transcription.call(url=url, audio_b64=audio_b64)


def run():
    password = st.text_input("Zugriffscode (siehe oben)")
    audio_file = st.file_uploader(
        "Datei auswählen", type=[".wav", ".mp3", ".flac", ".m4a", ".ogg"]
    )
    url = st.text_input(
        "URL (e.g. YouTube video, Dropbox file, etc.)",
        value="",
    )
    # https://www.youtube.com/watch?v=pLAaQO1iPz0
    submit_button = st.button(
        label="Transkribieren", disabled=(not audio_file and not url)
    )

    if audio_file is not None:
        st.audio(audio_file)
        cutoff = None if password == "" else 60_000
        audio_file = AudioSegment.from_file(audio_file)[:cutoff]
        audio_b64 = base64.b64encode(audio_file.export().read()).decode("ascii")
    if url:
        st.video(url)

    if submit_button:
        audio_b64 = None
        transcription = transcribe(url, audio_b64)
        for seg in transcription["text"].split("\n\n"):
            st.write(seg)
        st.json(transcription)


if __name__ == "__main__":
    try:
        run()
    except Exception as e:
        logger.error(e)
        st.error(
            "Leider ist ein unerwarter Fehler aufgetreten. Ich könnte mir das Problem sofort ansehen, Sie erreichen mich unter [email protected]"
        )