import gradio as gr
import torch
from transformers import pipeline

# Check if a GPU is available and set the device
device = 0 if torch.cuda.is_available() else -1

# Load the ASR model using the Hugging Face pipeline
model_id = "riteshkr/quantized-whisper-large-v3"  # Update with your model path or ID
pipe = pipeline("automatic-speech-recognition", model=model_id, device=device)

# Define the transcription function with batching support
def transcribe_speech(filepath):
    # Adjust batch size based on device (smaller batch for CPU)
    batch_size = 16 if torch.cuda.is_available() else 4
    
    output = pipe(
        filepath,
        max_new_tokens=256,
        generate_kwargs={
            "task": "transcribe",
            "language": "english",
        },
        chunk_length_s=30,
        batch_size=batch_size,  # Dynamic batch size
    )
    return output["text"]

# Define the Gradio interface for microphone input
mic_transcribe = gr.Interface(
    fn=transcribe_speech,
    inputs=gr.Audio(sources="microphone", type="filepath"),
    outputs=gr.Textbox(),
)

# Define the Gradio interface for file upload input
file_transcribe = gr.Interface(
    fn=transcribe_speech,
    inputs=gr.Audio(sources="upload", type="filepath"),
    outputs=gr.Textbox(),
)

# Creating the tabbed layout using Blocks
demo = gr.Blocks()

with demo:
    gr.TabbedInterface(
        [mic_transcribe, file_transcribe],
        ["Transcribe Microphone", "Transcribe Audio File"],
    )

# Launch the app with debugging enabled
if __name__ == "__main__":
    demo.launch(debug=True, share=True)