app.py

import os
import tempfile

import numpy as np
import torch
from fastapi import FastAPI, File, UploadFile
from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline

os.environ["TRANSFORMERS_CACHE"] = "/app/cache"

app = FastAPI()

# Device configuration
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-turbo"
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)

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

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

@app.post("/transcribe/")
async def transcribe_audio(file: UploadFile = File(...)):
    try:
        # Save the uploaded file temporarily
        with tempfile.NamedTemporaryFile(delete=True, suffix=".wav") as temp_audio:
            temp_audio.write(await file.read())
            temp_audio.flush()

            # Transcribe the audio
            result = pipe(temp_audio.name, return_timestamps="word")

        return {"transcription": result["chunks"]}

    except Exception as e:
        return {"error": str(e)}