main.py

from fastapi import FastAPI, File, UploadFile, HTTPException
from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline
import torch
import uvicorn
import soundfile as sf
from fastapi.middleware.cors import CORSMiddleware
import os
import tempfile

# Initialize FastAPI
app = FastAPI()
app.add_middleware(
    CORSMiddleware,
    allow_origins=['*'],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

# Load the model and processor
device = "cuda:0" if torch.cuda.is_available() else "cpu"
torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32

model_id = "openai/whisper-large-v3"

# Check if model exists locally, otherwise download it
if not os.path.exists(f"./{model_id}"):
    model = AutoModelForSpeechSeq2Seq.from_pretrained(model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True)
    processor = AutoProcessor.from_pretrained(model_id)
else:
    model = AutoModelForSpeechSeq2Seq.from_pretrained(f"./{model_id}", torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True)
    processor = AutoProcessor.from_pretrained(f"./{model_id}")

model.to(device)

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

# API endpoint to upload audio and get the transcribed text
@app.post("/transcribe")
async def transcribe_audio(file: UploadFile = File(...)):
    try:
        # Create a temporary file to save the uploaded content
        with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as temp_file:
            temp_file.write(await file.read())
            temp_path = temp_file.name

        # Load the audio file using soundfile
        audio, sr = sf.read(temp_path)

        # Ensure the sample rate is 16000 Hz
        if sr != 16000:
            raise HTTPException(status_code=400, detail="Sample rate must be 16000 Hz.")

        # Pass the processed audio to the pipeline
        result = pipe(temp_path)

        # Remove the temp file after processing
        os.remove(temp_path)

        # Return the transcribed text
        return {"text": result["text"]}
    
    except Exception as e:
        # Clean up temp file in case of error
        if 'temp_path' in locals() and os.path.exists(temp_path):
            os.remove(temp_path)
        raise HTTPException(status_code=500, detail=f"Error occurred: {str(e)}")

@app.get("/")
async def root():
    return {"message": "Welcome to the speech-to-text API!"}

# Running FastAPI with Uvicorn
if __name__ == "__main__":
    uvicorn.run(app, host="0.0.0.0", port=8000)