app.py

import gradio as gr
import subprocess
import torch
from charts import spider_chart
from dictionaries import calculate_average, transform_dict
from icon import generate_icon
from transformers import pipeline
from timestamp import format_timestamp
from youtube import get_youtube_video_id

MODEL_NAME = "openai/whisper-medium"
BATCH_SIZE = 8

device = 0 if torch.cuda.is_available() else "cpu"

pipe = pipeline(
    task="automatic-speech-recognition",
    model=MODEL_NAME,
    chunk_length_s=30,
    device=device,
)

#Formating
title = "Whisper Demo: Transcribe Audio"
MODEL_NAME1 = "jpdiazpardo/whisper-tiny-metal"
description = ("Transcribe long-form audio inputs with the click of a button! Demo uses the"
        f" checkpoint [{MODEL_NAME1}](https://huggingface.co/{MODEL_NAME1}) and 🤗 Transformers to transcribe audio files"
        " of arbitrary length. Check some of the 'cool' examples below")

examples = [["https://www.youtube.com/watch?v=W72Lnz1n-jw&ab_channel=Whitechapel-Topic",None,None,
            "examples/When a Demon Defiles a Witch.wav",True, True],
            ["https://www.youtube.com/watch?v=BnO3Io0KOl4&ab_channel=MotionlessInWhite-Topic",None,None,
             "examples/Immaculate Misconception.wav",True, True]]

linkedin = generate_icon("linkedin")
github = generate_icon("github")
article = ("<div style='text-align: center; max-width:800px; margin:10px auto;'>"
            f"<p>{linkedin} <a href='https://www.linkedin.com/in/juanpablodiazp/' target='_blank'>Juan Pablo Díaz Pardo</a><br>"
            f"{github} <a href='https://github.com/jpdiazpardo' target='_blank'>jpdiazpardo</a></p>")

title = "Scream: Fine-Tuned Whisper model for automatic gutural speech recognition 🤟🤟🤟"

#-------------------------------------------------------------------------------------------------------------------------------

#Define classifier for sentiment analysis
classifier = pipeline("text-classification", model="j-hartmann/emotion-english-distilroberta-base", top_k=None)

def transcribe(*args):#file, return_timestamps, *kwargs):
    '''inputs: file, return_timestamps'''
    outputs = pipe(args[3], batch_size=BATCH_SIZE, generate_kwargs={"task": 'transcribe'}, return_timestamps=True)
    text = outputs["text"]
    timestamps = outputs["chunks"]

    #If return timestamps is True, return html text with timestamps format
    if args[4]==True:
      spider_text = [f"{chunk['text']}" for chunk in timestamps] #Text for spider chart without timestamps
      timestamps = [f"[{format_timestamp(chunk['timestamp'][0])} -> {format_timestamp(chunk['timestamp'][1])}] {chunk['text']}" for chunk in timestamps]

    else:
      timestamps = [f"{chunk['text']}" for chunk in timestamps]
      spider_text = timestamps

    text = "<br>".join(str(feature) for feature in timestamps)
    text = f"<h4>Transcription</h4><div style='overflow-y: scroll; height: 150px;'>{text}</div>"

    spider_text = "\n".join(str(feature) for feature in spider_text)
    trans_dict=[transform_dict(classifier.predict(t)[0]) for t in spider_text.split("\n")]
    av_dict = calculate_average(trans_dict)
    fig = spider_chart(av_dict)
    
    return args[3], text, fig, av_dict

def filter(choice):
    if choice=="YouTube":
        return yt_link.update(interactive=True), audio_input.update(interactive=False)
    elif choice == "Upload File":
        return yt_link.update(value=None,interactive=False), audio_input.update(interactive=True)
    else:
        return yt_link.update(interactive=False), audio_input.update(interactive=False)


embed_html = '<iframe src="https://www.youtube.com/embed/YOUTUBE_ID'\
              'title="YouTube video player" frameborder="0" allow="accelerometer;'\
              'autoplay; clipboard-write; encrypted-media; gyroscope;'\
              'picture-in-picture" allowfullscreen></iframe>'

def download(link):
    subprocess.run(['python3', 'youtubetowav.py', link])
    return thumbnail.update(value=embed_html.replace("YOUTUBE_ID",get_youtube_video_id(link)), visible=True)

def hide_sa(value):
  if value == True:
    return sa_plot.update(visible=True), sa_frequency.update(visible=True)
  else:
    return sa_plot.update(visible=False), sa_frequency.update(visible=False)
  
#Input components        
yt_link = gr.Textbox(value=None,label="YouTube link", info = "Optional: Copy and paste YouTube URL")
audio_input = gr.Audio(source="upload", type="filepath", label="Upload audio file for transcription")
download_button = gr.Button("Download")
thumbnail = gr.HTML(value=embed_html, visible=False)
sa_checkbox = gr.Checkbox(value=True, label="Sentiment analysis")

inputs = [yt_link, #0
          download_button, #1
          thumbnail, #2
          audio_input, #3
          gr.Checkbox(value=True, label="Return timestamps"), #4
          sa_checkbox] #5

#Ouput components
audio_out = gr.Audio(label="Processed Audio", type="filepath", info = "Vocals only") 
sa_plot = gr.Plot(label="Sentiment Analysis")
sa_frequency = gr.Label(label="Frequency")

outputs = [audio_out, gr.outputs.HTML("text"), sa_plot, sa_frequency]

with gr.Blocks() as demo:
  download_button.click(download, inputs=[yt_link], outputs=[thumbnail])
  sa_checkbox.change(hide_sa, inputs=[sa_checkbox], outputs=[sa_plot, sa_frequency])


  with gr.Column():
      gr.Interface(title = title, fn=transcribe, inputs = inputs, outputs = outputs, 
                   description=description, cache_examples=True, allow_flagging="never", article = article , examples=None)
            
demo.queue(concurrency_count=3)
demo.launch(debug = True)