app.py

import gradio as gr
import torch
import whisper
from transformers import pipeline

### ————————————————————————————————————————

title="Whisper to Emotion"

### ————————————————————————————————————————

whisper_model = whisper.load_model("small")

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

emotion_classifier = pipeline("text-classification",model='bhadresh-savani/distilbert-base-uncased-emotion')

def translate_and_classify(audio):
    
    print("""
    —
    Sending audio to Whisper ...
    —
    """)
    audio = whisper.load_audio(audio)
    audio = whisper.pad_or_trim(audio)
    
    mel = whisper.log_mel_spectrogram(audio).to(whisper_model.device)
    
    _, probs = whisper_model.detect_language(mel)
    
    transcript_options = whisper.DecodingOptions(task="transcribe", fp16 = False)
    translate_options = whisper.DecodingOptions(task="translate", fp16 = False)
    
    transcription = whisper.decode(whisper_model, mel, transcript_options)
    translation = whisper.decode(whisper_model, mel, translate_options)
    
    print("Language Spoken: " + transcription.language)
    print("Transcript: " + transcription.text)  
    print("Translated: " + translation.text)
    
    emotion = emotion_classifier(translation.text)
    detected_emotion = emotion[0]["label"]
    return transcription.text, detected_emotion

with gr.Blocks() as demo:
    gr.Markdown("""
                ## Emotion Detection From Speech with Whisper
            """)
    
    with gr.Row():
        with gr.Column():
            #gr.Markdown(""" ### Record audio """)
            audio_input = gr.Audio(label = 'Record Audio Input',source="microphone",type="filepath")
            with gr.Row():
                transcribe_audio = gr.Button('Transcribe')
                
        with gr.Column():
            with gr.Row():
                transcript_output = gr.Textbox(label="Transcription in the language you spoke", lines = 3)
                emotion_output = gr.Textbox(label = "Detected Emotion")
    
    transcribe_audio.click(translate_and_classify, inputs = audio_input, outputs = [transcript_output,emotion_output])
    
    gr.HTML('''
        <div class="footer">
                    <p>Model by <a href="https://github.com/openai/whisper" style="text-decoration: underline;" target="_blank">OpenAI</a> 
                    </p>
        </div>
        ''')
    
    
demo.launch()