import torch
import gradio as gr
from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline

# Set up GPU if available
device = "cuda:0" if torch.cuda.is_available() else "cpu"
torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32

# Load Whisper model
model_id = "openai/whisper-large-v3"

model = AutoModelForSpeechSeq2Seq.from_pretrained(
    model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True
).to(device)

processor = AutoProcessor.from_pretrained(model_id)

# Initialize Whisper ASR pipeline
pipe = pipeline(
    "automatic-speech-recognition",
    model=model,
    tokenizer=processor.tokenizer,
    feature_extractor=processor.feature_extractor,
    torch_dtype=torch_dtype,
    device=device,
)

# Function to transcribe audio
def transcribe(audio_file):
    if not audio_file:
        return "Error: No audio provided."

    # Run ASR pipeline on the WAV file
    result = pipe(audio_file)
    return result["text"]

# Create Gradio UI with WAV format
demo = gr.Interface(
    fn=transcribe,
    inputs=gr.Audio(sources=["microphone", "upload"], type="filepath", label="Record or Upload WAV Audio"),
    outputs=gr.Textbox(),
    title="Whisper ASR (Speech-to-Text)",
    description="Transcribe spoken words into text using OpenAI Whisper Large V3. Supports WAV format.",
    live=True,
)

# Launch Gradio app
demo.launch()