handler.py

from transformers import pipeline
import base64


class TTSHandler:
    def __init__(self):
        self.pipe = pipeline("text-to-speech", model="suno/bark")
    
    def preprocess(self, text):
        return {"inputs": text.strip()}
    
    def inference(self, inputs):
        return self.pipe(**inputs)

    def postprocess(self, audio):
        return {"audio": base64.b64encode(audio).decode("utf-8")}