lenML/ChatTTS-Forge

1import logging2import random3from typing import Any, Dict, List4 5import numpy as np6from lxml import etree7 8from modules.data import styles_mgr9from modules.speaker import speaker_mgr10 11logger = logging.getLogger(__name__)12 13 14def expand_spk(attrs: dict):15    input_spk = attrs.get("spk", "")16    if isinstance(input_spk, int):17        return18    if isinstance(input_spk, str) and input_spk.isdigit():19        attrs.update({"spk": int(input_spk)})20        return21    try:22        speaker = speaker_mgr.get_speaker(input_spk)23        attrs.update({"spk": speaker})24    except Exception as e:25        logger.error(f"apply style failed, {e}")26 27 28def expand_style(attrs: dict):29    if attrs.get("style", "") != "":30        try:31            params = styles_mgr.find_params_by_name(str(attrs["style"]))32            attrs.update(params)33        except Exception as e:34            logger.error(f"apply style failed, {e}")35 36 37def merge_prompt(attrs: dict, elem):38 39    def attr_num(attrs: Dict[str, Any], k: str, min_value: int, max_value: int):40        val = elem.get(k, attrs.get(k, ""))41        if val == "":42            return43        if val == "max":44            val = max_value45        if val == "min":46            val = min_value47        val = np.clip(int(val), min_value, max_value)48        if "prefix" not in attrs or attrs["prefix"] == None:49            attrs["prefix"] = ""50        attrs["prefix"] += " " + f"[{k}_{val}]"51 52    attr_num(attrs, "oral", 0, 9)53    attr_num(attrs, "speed", 0, 9)54    attr_num(attrs, "laugh", 0, 2)55    attr_num(attrs, "break", 0, 7)56 57 58def apply_random_seed(attrs: dict):59    seed = attrs.get("seed", "")60    if seed == "random" or seed == "rand":61        seed = random.randint(0, 2**32 - 1)62        attrs["seed"] = seed63        logger.info(f"random seed: {seed}")64