app.py

import gradio as gr
from deep_translator import GoogleTranslator
#import deepl
from transformers import pipeline
import os
from gtts import gTTS
from pytube import YouTube

pipe = pipeline(model="freeja/lab2-whisper-sv")

def transcribe_audio(audio,language):
    transcribed = pipe(audio)["text"]
    result = "Transcribed text\n"
    result += transcribed + "\n"
    #result += "Translated text\n"
    #trans_text = translate_audio(transcribed,language)
    #result += trans_text
    #text_to_speech(trans_text,language)
    #result += text_to_speech
    return transcribed


def translate_audio(text,language):
    #translate = deepl.Translator
    language_dict = {"English":"en","Spanish":"es","German":"de","French":"fr","Italian":"it"}
    lang = language_dict[language]
    translated_text = GoogleTranslator(source='sv', target=lang).translate(text)
    return translated_text

def text_to_speech(text,language):
    language_dict = {"English":"en","Spanish":"es","German":"de","French":"fr","Italian":"it"}
    lang = language_dict[language]
    gTTS(text,lang,slow=False)

def transcribe_video(URL):
    video = YouTube(URL)
    yt = video.streams.get_audio_only()
    yt.download()
    text = pipe(yt)["text"]
    return text

"""iface = gr.Interface(
    fn=transcribe_audio,
    inputs=[
        gr.Audio(source="microphone", type="filepath", label="Transcribe from Microphone"),
        gr.Dropdown(["English","Spanish","Dutch","French","Italian"], value="English", label="Translate to ")
    ],
    outputs="text",
    title="Whisper Small Swedish",
    description="Realtime demo for Swedish speech recognition with translation using a fine-tuned Whisper small model")"""

video_transcription = gr.Interface(
    fn = transcribe_video, 
    inputs = "text", 
    outputs = "text",
    title = "Whisper Small Swedish",
    description = "Realtime demo for Swedish speech recognition with translation using a fine-tuned Whisper small model"",
)

audio_transcription = gr.Interface(
    fn=transcribe_audio, 
    inputs=[
        gr.Audio(source="microphone", type="filepath", label="Transcribe from Microphone"),
        gr.Dropdown(["English","Spanish","Dutch","French","Italian"], value="English", label="Translate to ")
    ], 
    outputs="text",
    title="Whisper Small Swedish",
    description="Realtime demo for Swedish speech recognition with translation using a fine-tuned Whisper small model"",
)

iface = gr.TabbedInterface([audio_transcription, video_transcription], ["Transcribe Audio", "Transcribe Video"])

iface.launch()