Misjudgment of is_multi_lingual When Loading Multilingual Model via model_path

[TITC]

Story

first problem-without model_name

The logic in api.py to sentence a model whether is multi-lingual or not is through model_name.

TTS/TTS/api.py

Lines 109 to 110 in 29dede2

	if isinstance(self.model_name, str) and "xtts" in self.model_name:
	return True

second problem-with model_name

but the problem is if the model_name is given then the ModelManager will try to download the model to the output_prefix, which is default to /home/yhtao/.local/share/tts in my system.

tts = TTS(model_path="/mnt/f/model/XTTS-v2_cache/tts_models--multilingual--multi-dataset--xtts_v2/",
          config_path="/mnt/f/model/XTTS-v2_cache/tts_models--multilingual--multi-dataset--xtts_v2//config.json").to(device)

TTS/TTS/utils/generic_utils.py

Lines 129 to 148 in 29dede2

	def get_user_data_dir(appname):
	TTS_HOME = os.environ.get("TTS_HOME")
	XDG_DATA_HOME = os.environ.get("XDG_DATA_HOME")
	if TTS_HOME is not None:
	ans = Path(TTS_HOME).expanduser().resolve(strict=False)
	elif XDG_DATA_HOME is not None:
	ans = Path(XDG_DATA_HOME).expanduser().resolve(strict=False)
	elif sys.platform == "win32":
	import winreg # pylint: disable=import-outside-toplevel
	key = winreg.OpenKey(
	winreg.HKEY_CURRENT_USER, r"Software\Microsoft\Windows\CurrentVersion\Explorer\Shell Folders"
	)
	dir_, _ = winreg.QueryValueEx(key, "Local AppData")
	ans = Path(dir_).resolve(strict=False)
	elif sys.platform == "darwin":
	ans = Path("~/Library/Application Support/").expanduser()
	else:
	ans = Path.home().joinpath(".local/share")
	return ans.joinpath(appname)

of course, there is no model under /home/yhtao/.local/share/tts, and then the ModelManager re-download model to this path.

TTS/TTS/utils/manage.py

Lines 378 to 402 in 29dede2

	output_path = os.path.join(self.output_prefix, model_full_name)
	if os.path.exists(output_path):
	if md5sum is not None:
	md5sum_file = os.path.join(output_path, "hash.md5")
	if os.path.isfile(md5sum_file):
	with open(md5sum_file, mode="r") as f:
	if not f.read() == md5sum:
	print(f" > {model_name} has been updated, clearing model cache...")
	self.create_dir_and_download_model(model_name, model_item, output_path)
	else:
	print(f" > {model_name} is already downloaded.")
	else:
	print(f" > {model_name} has been updated, clearing model cache...")
	self.create_dir_and_download_model(model_name, model_item, output_path)
	# if the configs are different, redownload it
	# ToDo: we need a better way to handle it
	if "xtts" in model_name:
	try:
	self.check_if_configs_are_equal(model_name, model_item, output_path)
	except:
	pass
	else:
	print(f" > {model_name} is already downloaded.")
	else:
	self.create_dir_and_download_model(model_name, model_item, output_path)

Usage

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

model_path = "/mnt/f/model/XTTS-v2_cache/tts_models--multilingual--multi-dataset--xtts_v2/"

tts = TTS(model_path="/mnt/f/model/XTTS-v2_cache/tts_models--multilingual--multi-dataset--xtts_v2/",
          config_path="/mnt/f/model/XTTS-v2_cache/tts_models--multilingual--multi-dataset--xtts_v2//config.json").to(device)


print("tts.is_multi_lingual", tts.is_multi_lingual)

speaker_wav = "/mnt/c/Users/yh.tao/Downloads/灿烂千阳-录音-声源.mp3"
text = "是的，本质上这是一个成本和代价的问题：1、如果追查者拥有很大的权力，例如国际刑警组织或者是国际CERT组织，可以协调全球各运营商，通过官方渠道一级一级追踪回去，只要时间足够，是可以追查到最后一级的真实接入的，凯文米特尼克就是这么被下村勉抓到的。2、如果追查者拥有很高的黑客技术，一级一级反向破解回去，入侵跳板机器并且监听链接，也是可以追查到最后一级的真实接入。3、有些跳板/VPN/代理/肉机等等本身就是陷阱，用了的话等于自我暴露。有些加密通道已经不再安全，例如SSL和tor，过于信任这种技术也会导致轻易暴露。4、大数据分析和行为模式分析有可能可以定位到具体的人，哪怕并不清楚这么多层的链接究竟是怎么构建的。"
file_path = "/mnt/c/Users/yh.tao/Downloads/灿烂千阳-录音-clone-知乎回答.mp3"

tts.tts_to_file(text=text, speaker_wav=speaker_wav,
                language="zh-cn", file_path=file_path)

[CLAassistant]

All committers have signed the CLA.

[erogol]

I understand the issue but your solution assumes that the custom path has the model name as the last folder. It is not always true. Is there a different way to solve this? Though it is not obvious to me.

[TITC]

There is a parameter in config.json

"model": "xtts",

Is this should always be true? if it is then the model_name could be set with it if not given.

besides this way, we can modify is_multi_lingual logic, there is a language list in config.json, so is_multi_lingual is true when the number of the list elements more than 1

    "languages": [
        "en",
        "es",
        "fr",
        "de",
        "it",
        "pt",
        "pl",
        "tr",
        "ru",
        "nl",
        "cs",
        "ar",
        "zh-cn",
        "hu",
        "ko",
        "ja"
    ],