17 lines
624 B
Python
17 lines
624 B
Python
"""TTS backends: speak(items=[(text, out_wav)], seed) in the cloned narrator voice.
|
|
|
|
local — qwen-tts in this process on a CUDA GPU (any machine, the default);
|
|
media_node — an MCP server with generate_speech (the author's home GPU node; see media_node.py).
|
|
"""
|
|
|
|
|
|
def backend(cfg, ref_wav, ref_text):
|
|
name = cfg.get("tts", "backend", "local")
|
|
if name == "local":
|
|
from .local import LocalTTS
|
|
return LocalTTS(cfg, ref_wav, ref_text)
|
|
if name == "media_node":
|
|
from .media_node import MediaNodeTTS
|
|
return MediaNodeTTS(cfg)
|
|
raise SystemExit(f"unknown tts.backend {name!r}")
|