import gradio as gr
import os
import whisper

model = whisper.load_model("base")

def transcribe(url):
  os.system(f"yt-dlp -x {url} -o audio.m4a")
  result = model.transcribe("audio.m4a")
  return result['text']

with gr.Blocks() as demo:
  gr.HTML(
    """
      <p>
        Transcribes web videos to text using OpenAI's Whisper (base model).
        Works on many sites such as YouTube, Twitter, Reddit, Instagram, etc.
      </p>
    """
  )
  url = gr.Textbox(placeholder="Enter video link here...", label="")
  button = gr.Button("Transcribe!")
  output = gr.Textbox(label="")
  button.click(transcribe, inputs=[url], outputs=[output])
  
demo.launch()