kevinzhow/style-bert-vits2-gguf

Style-Bert-VITS2 GGUF Models

GGUF runtime bundle for the myoland.tts Style-Bert-VITS2 GGML backend and the TTS.cpp Style-Bert-VITS2 model servers.

These files are conversion artifacts for local GGML inference. They are not new trained models. Use each artifact under the original upstream model or voice terms, including the attribution requirements from the source model pages.

Layout

• voices/*.gguf: Style-Bert-VITS2 acoustic models used by the TTS.cpp decoder server.
• frontend/style-bert-vits2-jp-bert.gguf: Japanese JP-BERT feature model used by the Style-Bert-VITS2 GGML frontend path.
• MANIFEST.sha256: SHA-256 checksums for uploaded GGUF files.

Source Models

GGUF artifact	Upstream source	Notes
frontend/style-bert-vits2-jp-bert.gguf	ku-nlp/deberta-v2-large-japanese-char-wwm	Japanese DeBERTa / JP-BERT feature model converted for the TTS.cpp Style-Bert-VITS2 JP-BERT server. Tokenizer files come from the same upstream model.
voices/jvnv-F1-jp-full-sdp.gguf	litagin/style_bert_vits2_jvnv	Converted from jvnv-F1-jp_e160_s14000.safetensors plus the matching Style-Bert-VITS2 config and style vectors.
voices/jvnv-F2-jp-full-sdp.gguf	litagin/style_bert_vits2_jvnv	Converted from jvnv-F2_e166_s20000.safetensors plus the matching Style-Bert-VITS2 config and style vectors.
voices/jvnv-M1-jp-full-sdp.gguf	litagin/style_bert_vits2_jvnv	Converted from jvnv-M1-jp_e158_s14000.safetensors plus the matching Style-Bert-VITS2 config and style vectors.
voices/jvnv-M2-jp-full-sdp.gguf	litagin/style_bert_vits2_jvnv	Converted from jvnv-M2-jp_e159_s17000.safetensors plus the matching Style-Bert-VITS2 config and style vectors.
voices/amitaro-full-sdp.gguf	litagin/sbv2_amitaro	Converted from the amitaro Style-Bert-VITS2 assets. The source terms require visible credit such as Style-BertVITS2 model: あみたろ、あみたろの声素材工房 (https://amitaro.net/).
voices/koharune-ami-full-sdp.gguf	litagin/sbv2_koharune_ami	Converted from the koharune-ami Style-Bert-VITS2 assets. The source terms require visible credit such as Style-BertVITS2モデル: 小春音アミ、あみたろの声素材工房 (https://amitaro.net/).
voices/mao-full-sdp.gguf	AivisHub まお	Downloaded as an AIVM Safetensors package and converted to the local Style-Bert-VITS2 asset layout before GGUF conversion. AIVM model UUID: a59cb814-0083-4369-8542-f51a29e72af7; package SHA-256: 1c8ec3160c80b14eb25e216a981d066a029f41f5594862ea4b5f063a59575846.
voices/michinoku-airi-full-sdp.gguf	AivisHub みちのくあいり	Downloaded as an AIVM Safetensors package and converted to the local Style-Bert-VITS2 asset layout before GGUF conversion. AIVM model UUID: 1b2830f4-8cf1-4184-a0d9-3a1bace3a844; package SHA-256: cd5392d4996121b37b2cd2372b5658da967b98f3670796868284bc8b9c58df21.
voices/chu2-full-sdp.gguf	AivisHub 中2	Downloaded as an AIVM Safetensors package and converted to the local Style-Bert-VITS2 asset layout before GGUF conversion. AIVM model UUID: 9107b8b6-1ed1-43f5-bebe-0de4df4d229d; package SHA-256: 1a959653a7844dc85273b3c5b1b703e5883dbdf60d2822ca4e77133124fefd60.
voices/nise-full-sdp.gguf	AivisHub にせ	Downloaded as an AIVM Safetensors package and converted to the local Style-Bert-VITS2 asset layout before GGUF conversion. AIVM model UUID: 6d11c6c2-f4a4-4435-887e-23dd60f8b8dd; package SHA-256: ed38a9a6c5e08c7a25b6ea1db0032d365530637896f874aaa55e627b1e1cf91b.
voices/kanon-full-sdp.gguf	AivisHub 花音	Downloaded as an AIVM Safetensors package and converted to the local Style-Bert-VITS2 asset layout before GGUF conversion. AIVM model UUID: a670e6b8-0852-45b2-8704-1bc9862f2fe6; package SHA-256: 7f0f744a37cb89db46725c3ff6cf14668e9a99f41e232266d14ea0ead67b5d53.
voices/mai-full-sdp.gguf	AivisHub まい	Downloaded as an AIVM Safetensors package and converted to the local Style-Bert-VITS2 asset layout before GGUF conversion. AIVM model UUID: e9339137-2ae3-4d41-9394-fb757a7e61e6; package SHA-256: 7ab31ef87b9c1c1f6fce8c9a0f6dea7a812875c70598048339483b4540559db6.
voices/runa-full-sdp.gguf	AivisHub るな	Downloaded as an AIVM Safetensors package and converted to the local Style-Bert-VITS2 asset layout before GGUF conversion. AIVM model UUID: 4f281e78-eba6-495a-8e50-5c322d02b5b1; package SHA-256: b1ef9a21f3beb63981ee34e799684379084c14a20aad2b4fdfba5f7f93bc3417.

Conversion Notes

Voice artifacts were converted from Style-Bert-VITS2 config.json, style_vectors.npy, and .safetensors model assets into the TTS.cpp GGUF format. AIVM voices were first unpacked into the same Style-Bert-VITS2 asset layout, then converted to GGUF.

The full-sdp suffix is the local TTS.cpp naming convention for the converted Style-Bert-VITS2 model variant used by the GGML frontend path.

Runtime Use

The expected local layout is:

tmp/style-bert-vits2-gguf/
  frontend/style-bert-vits2-jp-bert.gguf
  voices/*.gguf

Example download:

hf download kevinzhow/style-bert-vits2-gguf \
  voices/jvnv-F1-jp-full-sdp.gguf \
  frontend/style-bert-vits2-jp-bert.gguf \
  --local-dir /path/to/TTS.cpp/tmp/style-bert-vits2-gguf