WildAi
9574c1bc87
Borrowed from VoxCPM (HF download stack, settings, routes support) and VibeVoice (registry, patcher, device/dtype/ attention). Deletes chatterbox_handler.py.
121 lines
4.4 KiB
Python
121 lines
4.4 KiB
Python
"""Frozen constants for the Chatterbox node pack.
|
|
|
|
These four values were defined in ``modules/chatterbox_handler.py``, deleted by
|
|
S3.2, and are recorded in ``tests/frozen_widget_snapshot.json``. They are the
|
|
compatibility contract: saved workflows were built against the default pack
|
|
name and the on-disk layout under ``models/tts/chatterbox``, so neither may
|
|
drift. ``tests/test_move_parity.py::TestCarriedForwardConstants`` compares this
|
|
module against that record.
|
|
|
|
Nothing here imports ``folder_paths``, ``comfy.*`` or ``server``: it is a leaf
|
|
module every other module may import from freely.
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
import os
|
|
|
|
__all__ = [
|
|
"CHATTERBOX_MODEL_SUBDIR",
|
|
"CHATTERBOX_REPO_ID",
|
|
"CHATTERBOX_FILES_TO_DOWNLOAD",
|
|
"DEFAULT_MODEL_PACK_NAME",
|
|
"MTL_LANGUAGES",
|
|
"MTL_LANGUAGE_CODES",
|
|
"MTL_LANGUAGE_LOOKUP",
|
|
"MTL_T3_MODELS",
|
|
"MTL_DEFAULT_T3_MODEL",
|
|
]
|
|
|
|
# Where Chatterbox packs live, relative to a ComfyUI models root. Built with
|
|
# ``os.path.join`` so it renders with the platform separator; the snapshot
|
|
# records it platform-neutrally as the two parts below.
|
|
CHATTERBOX_MODEL_SUBDIR = os.path.join("tts", "chatterbox")
|
|
|
|
# The upstream Hugging Face repository the default voice pack is fetched from.
|
|
CHATTERBOX_REPO_ID = "ResembleAI/chatterbox"
|
|
|
|
# The complete file list of a default pack. Order is the order
|
|
# ``HFProgressTracker`` downloads them in, so it is load-bearing: the voice
|
|
# encoder is the slowest piece and the frontend renders progress per file.
|
|
CHATTERBOX_FILES_TO_DOWNLOAD = [
|
|
"ve.safetensors",
|
|
"t3_cfg.safetensors",
|
|
"s3gen.safetensors",
|
|
"tokenizer.json",
|
|
"conds.pt",
|
|
]
|
|
|
|
# The pack a saved workflow names when its combo widget carries no value.
|
|
DEFAULT_MODEL_PACK_NAME = "resembleai_default_voice"
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# Multilingual (S8.1)
|
|
# ---------------------------------------------------------------------------
|
|
#
|
|
# A copy of the vendored ``chatterbox.mtl_tts.SUPPORTED_LANGUAGES``, and it has to
|
|
# be a copy: the vendored module imports the whole model stack (librosa,
|
|
# transformers, safetensors, huggingface_hub), so importing it to build a Combo's
|
|
# option list would cost a ComfyUI startup. It also has to match *exactly* --
|
|
# a language the dropdown offers but the tokenizer rejects is a node that fails
|
|
# at generation time with a message the user cannot act on. The two are pinned
|
|
# against each other by ``tests/test_mtl_node.py``, which reads the real
|
|
# vendored dict in a subprocess and compares it with the table here.
|
|
#
|
|
# 23 codes. There is no Vietnamese: the Vietnamese and Egyptian-Arabic models are
|
|
# community finetunes, surfaced as ordinary local directories (S8.8) rather than
|
|
# as languages upstream does not define.
|
|
MTL_LANGUAGES: dict[str, str] = {
|
|
"ar": "Arabic",
|
|
"da": "Danish",
|
|
"de": "German",
|
|
"el": "Greek",
|
|
"en": "English",
|
|
"es": "Spanish",
|
|
"fi": "Finnish",
|
|
"fr": "French",
|
|
"he": "Hebrew",
|
|
"hi": "Hindi",
|
|
"it": "Italian",
|
|
"ja": "Japanese",
|
|
"ko": "Korean",
|
|
"ms": "Malay",
|
|
"nl": "Dutch",
|
|
"no": "Norwegian",
|
|
"pl": "Polish",
|
|
"pt": "Portuguese",
|
|
"ru": "Russian",
|
|
"sv": "Swedish",
|
|
"sw": "Swahili",
|
|
"tr": "Turkish",
|
|
"zh": "Chinese",
|
|
}
|
|
|
|
#: The 23 codes in the order the Combo offers them.
|
|
MTL_LANGUAGE_CODES: tuple[str, ...] = tuple(sorted(MTL_LANGUAGES))
|
|
|
|
#: Lower-cased code *or* display name -> code, so ``[German:Alice]`` and
|
|
#: ``[de:Alice]`` resolve to the same thing.
|
|
MTL_LANGUAGE_LOOKUP: dict[str, str] = {
|
|
**{code: code for code in MTL_LANGUAGES},
|
|
**{name.lower(): code for code, name in MTL_LANGUAGES.items()},
|
|
}
|
|
|
|
#: ``model_version`` Combo value -> T3 checkpoint filename.
|
|
#:
|
|
#: The Combo shows the *resolved filename*, not a bare ``v3`` label, because the
|
|
#: third-party implementations disagree about what "v3" even is and a user
|
|
#: comparing a graph against a Hugging Face listing needs the file name. Upstream
|
|
#: still resolves ``None`` to v2 (``DEFAULT_MULTILINGUAL_T3_MODEL``); we default
|
|
#: to **v3**, and the docstring on the node says why that differs.
|
|
MTL_T3_MODELS: dict[str, str] = {
|
|
"v3": "t3_mtl23ls_v3.safetensors",
|
|
"v2": "t3_mtl23ls_v2.safetensors",
|
|
"v1": "t3_23lang.safetensors",
|
|
}
|
|
|
|
#: The version the multilingual node starts on. v3, deliberately not upstream's
|
|
#: v2: the reference corpus records v3 as the recommended checkpoint and notes
|
|
#: the two reference implementations disagree on the default.
|
|
MTL_DEFAULT_T3_MODEL = "v3"
|