app.py

from transformers import pipeline
import gradio as gr
from pytube import YouTube
import os

pipe = pipeline(model="CsanadT/whisper_small_sv")

def transcribe_live(audio):
    text = pipe(audio)["text"]
    return text

def transcribe_url(url):
    youtube = YouTube(str(url))
    audio = youtube.streams.filter(only_audio=True).first().download('yt_video')
    text = pipe(audio)["text"]
    return text

def transcribe_file(audio):
    rate, y = audio
    text = pipe(y)["text"]
    return text

url_demo = gr.Interface(
    fn = transcribe_url, 
    inputs = "text", 
    outputs = "text",
    title = "Swedish Whisper",
    description = "Fine-tuned Whisper model for swedish audio transcription",
)

voice_demo = gr.Interface(
    fn=transcribe_live, 
    inputs=gr.Audio(source="microphone", type="filepath"), 
    outputs="text",
    title="Whisper Swedish",
    description="Fine-tuned Whisper model for swedish audio transcription",
)

file_demo = gr.Interface(
    fn = transcribe_file,
    inputs=gr.Audio(file_count="single"),
    outputs="text",
    title="Swedish Whisper",
    description="Fine-tuned Whisper model for swedish audio transcription",
)
demo = gr.TabbedInterface([url_demo, voice_demo, file_demo], ["YouTube video transciption", "Live audio to Text", "Transcribe a file"])

demo.launch()