main.py

import torch
import torchaudio
import numpy as np
from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline
from datasets import load_dataset
from googletrans import Translator
from fastapi import FastAPI, File, UploadFile, HTTPException
from fastapi.responses import JSONResponse
from pathlib import Path

app = FastAPI()

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"

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

processor = AutoProcessor.from_pretrained(model_id)

pipe = pipeline(
    "automatic-speech-recognition",
    model=model,
    tokenizer=processor.tokenizer,
    feature_extractor=processor.feature_extractor,
    max_new_tokens=256,
    chunk_length_s=30,
    batch_size=16,
    return_timestamps=True,
    torch_dtype=torch_dtype,
    device=device,
)

dataset = load_dataset("distil-whisper/librispeech_long", "clean", split="validation")

@app.post("/voice_recognition")
async def process_audio(file: UploadFile = File(...)):
    try:
        # File
        save_directory = Path("/home/user")
        save_directory.mkdir(parents=True, exist_ok=True)
        file_location = save_directory / file.filename

        with open(file_location, "wb") as saved_file:
            content = await file.read()
            saved_file.write(content)

        # Load audio file using torchaudio
        waveform, sample_rate = torchaudio.load(file_location)

        # Ensure the waveform is a 1D array
        waveform = waveform[0] if waveform.size(0) > 1 else waveform

        # ASR
        original = pipe(waveform.numpy())
        original_version = original[0]["text"]

        # EN
        result = pipe(waveform.numpy(), generate_kwargs={"task": "translate"})
        hasil = result[0]["text"]

        # ID
        detect = detect_google(hasil)
        id_ver = translate_google(hasil, f"{detect}", "ID")

        # Additional modifications
        id_ver = modify_text(id_ver)

        return JSONResponse(content={"response": {"jp_text": original_version, "en_text": hasil, "id_text": id_ver}}, status_code=200)

    except Exception as e:
        return HTTPException(status_code=500, detail=f"Error: {e}")

def detect_google(text):
    try:
        translator = Translator()
        detected_lang = translator.detect(text)
        return detected_lang.lang.upper()
    except Exception as e:
        print(f"Error detect: {e}")
        return None

def translate_google(text, source, target):
    try:
        translator = Translator()
        translated_text = translator.translate(text, src=source, dest=target)
        return translated_text.text
    except Exception as e:
        print(f"Error translate: {e}")
        return None

def modify_text(text):
    # Additional modifications, case-sensitive
    replacements = {
        "Tuan": "Master",
        "tuan": "Master",
        "Guru": "Master",
        "guru": "Master",
        "Monica": "Monika",
        "monica": "Monika",
    }

    for original, replacement in replacements.items():
        text = text.replace(original, replacement)

    return text