import asyncio import datetime import logging import os import time import traceback import tempfile from concurrent.futures import ThreadPoolExecutor import edge_tts import librosa import torch from fairseq import checkpoint_utils import uuid from config import Config from lib.infer_pack.models import ( SynthesizerTrnMs256NSFsid, SynthesizerTrnMs256NSFsid_nono, SynthesizerTrnMs768NSFsid, SynthesizerTrnMs768NSFsid_nono, ) from rmvpe import RMVPE from vc_infer_pipeline import VC # Set logging levels logging.getLogger("fairseq").setLevel(logging.WARNING) logging.getLogger("numba").setLevel(logging.WARNING) logging.getLogger("markdown_it").setLevel(logging.WARNING) logging.getLogger("urllib3").setLevel(logging.WARNING) logging.getLogger("matplotlib").setLevel(logging.WARNING) limitation = os.getenv("SYSTEM") == "spaces" config = Config() # Edge TTS tts_voice_list = asyncio.get_event_loop().run_until_complete(edge_tts.list_voices()) tts_voices = ["mn-MN-BataaNeural", "mn-MN-YesuiNeural"] # Specific voices # RVC models model_root = "weights" models = [d for d in os.listdir(model_root) if os.path.isdir(f"{model_root}/{d}")] models.sort() def get_unique_filename(extension): return f"{uuid.uuid4()}.{extension}" def model_data(model_name): pth_path = [ f"{model_root}/{model_name}/{f}" for f in os.listdir(f"{model_root}/{model_name}") if f.endswith(".pth") ][0] print(f"Loading {pth_path}") cpt = torch.load(pth_path, map_location="cpu") tgt_sr = cpt["config"][-1] cpt["config"][-3] = cpt["weight"]["emb_g.weight"].shape[0] # n_spk if_f0 = cpt.get("f0", 1) version = cpt.get("version", "v1") if version == "v1": if if_f0 == 1: net_g = SynthesizerTrnMs256NSFsid(*cpt["config"], is_half=config.is_half) else: net_g = SynthesizerTrnMs256NSFsid_nono(*cpt["config"]) elif version == "v2": if if_f0 == 1: net_g = SynthesizerTrnMs768NSFsid(*cpt["config"], is_half=config.is_half) else: net_g = SynthesizerTrnMs768NSFsid_nono(*cpt["config"]) else: raise ValueError("Unknown version") del net_g.enc_q net_g.load_state_dict(cpt["weight"], strict=False) print("Model loaded") net_g.eval().to(config.device) if config.is_half: net_g = net_g.half() else: net_g = net_g.float() vc = VC(tgt_sr, config) index_files = [ f"{model_root}/{model_name}/{f}" for f in os.listdir(f"{model_root}/{model_name}") if f.endswith(".index") ] if len(index_files) == 0: print("No index file found") index_file = "" else: index_file = index_files[0] print(f"Index file found: {index_file}") return tgt_sr, net_g, vc, version, index_file, if_f0 def load_hubert(): models, _, _ = checkpoint_utils.load_model_ensemble_and_task( ["hubert_base.pt"], suffix="", ) hubert_model = models[0] hubert_model = hubert_model.to(config.device) if config.is_half: hubert_model = hubert_model.half() else: hubert_model = hubert_model.float() return hubert_model.eval() def get_model_names(): model_root = "weights" return [d for d in os.listdir(model_root) if os.path.isdir(f"{model_root}/{d}")] def run_async_in_thread(fn, *args): loop = asyncio.new_event_loop() asyncio.set_event_loop(loop) result = loop.run_until_complete(fn(*args)) loop.close() return result async def tts( model_name, tts_text, tts_voice, index_rate, use_uploaded_voice, uploaded_voice, ): # ... (keep the existing implementation) voice_mapping = { "Mongolian Male": "mn-MN-BataaNeural", "Mongolian Female": "mn-MN-YesuiNeural" } hubert_model = load_hubert() rmvpe_model = RMVPE("rmvpe.pt", config.is_half, config.device) class TTSProcessor: def __init__(self, config): self.config = config self.executor = ThreadPoolExecutor(max_workers=config.n_cpu) self.semaphore = asyncio.Semaphore(10) # Limit to 10 concurrent tasks self.queue = asyncio.Queue() self.is_processing = False async def tts(self, model_name, tts_text, tts_voice, index_rate, use_uploaded_voice, uploaded_voice): task = asyncio.create_task(self._tts_task(model_name, tts_text, tts_voice, index_rate, use_uploaded_voice, uploaded_voice)) await self.queue.put(task) if not self.is_processing: asyncio.create_task(self._process_queue()) return await task async def _tts_task(self, model_name, tts_text, tts_voice, index_rate, use_uploaded_voice, uploaded_voice): async with self.semaphore: return await tts(model_name, tts_text, tts_voice, index_rate, use_uploaded_voice, uploaded_voice) async def _process_queue(self): self.is_processing = True while not self.queue.empty(): task = await self.queue.get() try: await task except Exception as e: print(f"Error processing TTS task: {str(e)}") finally: self.queue.task_done() self.is_processing = False # Initialize the TTSProcessor tts_processor = TTSProcessor(config) async def parallel_tts(tasks): return await asyncio.gather(*(tts_processor.tts(*task) for task in tasks)) def parallel_tts_wrapper(tasks): loop = asyncio.get_event_loop() return loop.run_until_complete(parallel_tts(tasks))