app.py

# import gradio as gr
# from transformers import pipeline

# # Load an audio classification pipeline (you can replace it with your desired model)
# audio_classification_pipeline = pipeline(task="automatic-speech-recognition", model="openai/whisper-small")

# def predict(input_audio):
#     # Use the audio classification pipeline to predict
#     transcription = audio_classification_pipeline(input_audio)["text"]
#     return transcription

# # Gradio interface setup for audio classification
# gradio_app = gr.Interface(
#     predict,
#     inputs=gr.Audio(label="Upload or Record Audio", type="numpy"),  # Audio input
#     outputs=gr.Textbox(label="Transcription Result"),               # Output is transcription text
#     title="Audio Classification: Speech-to-Text",
#     description="Upload or record an audio file to transcribe it using the Whisper model."
# )

# # Launch the app
# if __name__ == "__main__":
#     gradio_app.launch()


import gradio as gr
from transformers import pipeline

# Load the Whisper speech-to-text pipeline
audio_classification_pipeline = pipeline(task="automatic-speech-recognition", model="openai/whisper-small")

def predict(input_audio):
    if input_audio is None:
        return "Please upload an audio file."

    # Gradio audio input returns a tuple: (audio array, sample rate)
    audio_data, sample_rate = input_audio

    # Whisper expects a dictionary with keys 'array' and 'sampling_rate'
    transcription = audio_classification_pipeline({"array": audio_data, "sampling_rate": sample_rate})["text"]
    
    return transcription

# Gradio interface setup for audio classification
gradio_app = gr.Interface(
    fn=predict,
    inputs=gr.Audio(label="Upload or Record Audio", type="numpy"),  # Audio input
    outputs=gr.Textbox(label="Transcription Result"),               # Output is transcription text
    title="Audio Classification: Speech-to-Text",
    description="Upload or record an audio file to transcribe it using the Whisper model."
)

# Launch the app
if __name__ == "__main__":
    gradio_app.launch()