File size: 1,148 Bytes
3e55f57 7a8a532 f067669 3e55f57 7a8a532 3e55f57 4f9c202 7a8a532 62f977c 7a8a532 4f9c202 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 | from typing import Dict, List, Any
from transformers import AutoProcessor, AutoModel
import scipy.io.wavfile # Assuming WAV output format
class EndpointHandler:
def __init__(self, path=""):
self.processor = AutoProcessor.from_pretrained("suno/bark")
self.model = AutoModel.from_pretrained("suno/bark")
def __call__(self, data: Dict[str, Any]) -> Dict[str, Any]:
try:
text_prompt = data.get("inputs")
if not text_prompt:
raise ValueError("Missing required 'inputs' field in request data.")
inputs = self.processor(text=[text_prompt], return_tensors="pt")
speech_values = self.model.generate(**inputs, do_sample=True)
# Assuming model returns audio as NumPy array
audio_data = speech_values[0].numpy()
sampling_rate = 22050 # Adjust as needed based on model documentation
# Return audio data as a byte string
audio_bytes = audio_data.tobytes()
return {"audio": audio_bytes, "sampling_rate": sampling_rate}
except Exception as e:
return {"error": str(e)}
|