Skip to content

License issue for Japanese g2p in MaskGCT #446

@litagin02

Description

@litagin02

Hello, and thanks for nice toolkit!

When I tried to use MaskGCT and looked some codes in it, I realized that it seems that a code on Japanese g2p was borrowed from my Style-Bert-VITS2 repo, which is AGPL-v3.0 license (inherited from Bert-VITS2)

Concretely, most parts of the following file

# Copyright (c) 2024 Amphion.
#
# This source code is licensed under the MIT license found in the
# LICENSE file in the root directory of this source tree.

is from
https://github.com/litagin02/Style-Bert-VITS2/blob/master/style_bert_vits2/nlp/japanese/g2p.py
and other files in that directory, but the file maskgct/g2p/g2p/japanese.py is licensed as MIT under the name of Amphion.

The fact that this file comes from my repo is clear from e.g. the comment

# # "~": "-", # これは長音記号「ー」として扱うよう変更

https://github.com/litagin02/Style-Bert-VITS2/blob/1d7a7a0d484d56eebe99c5fd88a9ba52091b87bd/style_bert_vits2/nlp/japanese/normalizer.py#L57

def g2phone_tone_wo_punct(prosodies) -> list[tuple[str, int]]:
"""
テキストに対して、音素とアクセント(0か1)のペアのリストを返す。
ただし「!」「.」「?」等の非音素記号(punctuation)は全て消える(ポーズ記号も残さない)。
非音素記号を含める処理は`align_tones()`で行われる。
また「っ」は「cl」でなく「q」に変換される(「ん」は「N」のまま)。
例: "こんにちは、世界ー。。元気?!" →
[('k', 0), ('o', 0), ('N', 1), ('n', 1), ('i', 1), ('ch', 1), ('i', 1), ('w', 1), ('a', 1), ('s', 1), ('e', 1), ('k', 0), ('a', 0), ('i', 0), ('i', 0), ('g', 1), ('e', 1), ('N', 0), ('k', 0), ('i', 0)]
"""

https://github.com/litagin02/Style-Bert-VITS2/blob/1d7a7a0d484d56eebe99c5fd88a9ba52091b87bd/style_bert_vits2/nlp/japanese/g2p.py#L374-L381

def kata2phoneme_list(text: str) -> list[str]:
"""
原則カタカナの`text`を受け取り、それをそのままいじらずに音素記号のリストに変換。
注意点:
- punctuationが来た場合(punctuationが1文字の場合がありうる)、処理せず1文字のリストを返す
- 冒頭に続く「ー」はそのまま「ー」のままにする(`handle_long()`で処理される)
- 文中の「ー」は前の音素記号の最後の音素記号に変換される。
例:
`ーーソーナノカーー` → ["ー", "ー", "s", "o", "o", "n", "a", "n", "o", "k", "a", "a", "a"]
`?` → ["?"]
"""

https://github.com/litagin02/Style-Bert-VITS2/blob/1d7a7a0d484d56eebe99c5fd88a9ba52091b87bd/style_bert_vits2/nlp/japanese/g2p.py#L609-L619

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions