Skip to content
Other

Kimodo-SOMA-RP-v1.2

3D-анимация по тексту и ограничениям позы

2 min readupdated

Kimodo-SOMA-RP-v1.2 — обновлённая модель NVIDIA семейства Kimodo (Kinematic Motion Diffusion). Это диффузионная two-stage Transformer-модель на 282M параметров, которая генерирует 3D-скелетную анимацию тела по текстовому промпту и/или ограничениям движения. Обучена на 30-суставном скелете SOMA и проприетарном датасете Bones Rigplay.

Ключевые особенности

  • Генерирует 3D-скелетную анимацию по тексту и ограничениям: позам всего тела, позициям концевых суставов, траекториям и вейпоинтам.
  • Архитектура — диффузионная two-stage Transformer; 282M параметров, выход 30 fps.
  • В v1.2 исправлены train-сплиты: убраны тестовые последовательности BONES-SEED, попавшие в обучение v1.1. Меньше foot skating, точность ограничений на уровне v1.1, но R-precision/FID немного ниже.
  • Разрешено коммерческое использование по NVIDIA Open Model License: демонстрации для гуманоидных роботов, digital twin, синтетические данные, игры и медиа.
  • Работает на NVIDIA GPU поколений Ampere, Blackwell и Lovelace через PyTorch; поддерживаются Linux и Windows.

Пример

Промпт: «A person walks slowly with shuffled steps».
Ограничения: ключевые позы, позиции суставов и траектория движения.
Результат: до 300 кадров скелетной анимации при 30 fps — максимум 10 секунд.

ПодробнееКонкретных цифр бенчмарка в карточке нет — есть отсылка к таблицам Motion Generation Benchmark.
Выход: root translation и joint rotations, 30 fps, до 10 с.
Тестировали на RTX 3090/4090/5090, A100, L40S, L4, RTX 6000 Ada и RTX A6000.
Ограничения: возможен foot skating, модель не всегда следует тексту, не учитывает объекты сцены и выдаёт движение для одного скелета персонажа.
Discuss Telegram 0 19

Will it run?

Runs in F32

Fits without quantisation.

  • F32weights 1.1 GB · ready files1.7 GB
  • BF16weights 0.5 GB1.1 GB
  • FP8 / INT8recommendedweights 0.3 GB0.8 GB
  • 4-bitweights 0.1 GB0.7 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · F32 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

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.

    # Liserk/Kimodo-SOMA-RP-v1.2
    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 Liserk/Kimodo-SOMA-RP-v1.2 --local-dir ~/models/Kimodo-SOMA-RP-v1.2
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to Kimodo-SOMA-RP-v1.2
  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 Kimodo-SOMA-RP-v1.2 (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

Liserk/Kimodo-SOMA-RP-v1.2
Author
Liserk
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

190

Ask AI about this model

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

Comments