Skip to content
Audio

KurisuMakiseTTS

голос Макисэ Курису на GPT-SoVITS v5

1 min readupdated

Фанатский TTS-файнтюн от FishelKET на базе GPT-SoVITS v5dev и v5turbo. Японский — основной язык, китайский — кросс-языковой; выход 48 kHz. Обучен на 972 фрагментах (67,8 мин), для обучения отобрано 754 фрагмента (60,7 мин).

Ключевые особенности

  • Два варианта: v5dev — лучше качество; v5turbo — быстрее, но чуть уступает из-за 4-шаговой дистилляции.
  • SoVITS LoRA rank 32 + полный файнтюн GPT; обе версии используют общий GPT-чекпоинт.
  • GPU RTF: ~1.08 у dev и ~0.74–0.82 у turbo.
  • CPU RTF: ~32 у dev (практически непригодно) и ~3.2 у turbo; 4-секундная фраза — ~131 с против ~12 с.
  • Нужна референсная аудиозапись 5–10 секунд: она задаёт в основном интонацию, а не сходство голоса.

Пример

curl http://127.0.0.1:9880/tts -H "Content-Type: application/json" -d '{
  "text": "早上好,我是牧濑红莉栖。",
  "text_lang": "zh",
  "ref_audio_path": "ref_audio/ref1.wav",
  "prompt_text": "あなたに会いに来たの、岡部凛太郎さん。じゃなくて…",
  "prompt_lang": "ja",
  "media_type": "wav"
}' -o out.wav
ПодробнееВеса — Apache-2.0, но права на персонажа и голос принадлежат 5pb. / MAGES. / Nitroplus и актрисе озвучивания; для коммерческого использования голоса и персонажа нужно отдельное разрешение.
Обучение: SoVITS 12 эпох LoRA(r32) + GPT 15 эпох, около 8 и 6 минут на RTX PRO 6000.
Ограничение: около 73% обучающих данных — рациональная речь, эмоциональных образцов меньше.
Discuss Telegram 2

Will it run?

Working out the memory needs

The bot is still collecting this model's weight sizes; the estimate will appear here in a few minutes. Meanwhile you can rent a GPU with headroom.

Massed ComputeOpen Massed Compute

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # FishelKET/KurisuMakiseTTS
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download FishelKET/KurisuMakiseTTS --local-dir ~/models/KurisuMakiseTTS
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to KurisuMakiseTTS
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of KurisuMakiseTTS (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

FishelKET/KurisuMakiseTTS
Author
FishelKET
Category
Audio
Released
Posted

On NeyroLearn

Hugging Face

02

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments