Skip to content
Other

H3 Singularity ActionGraft

быстрее драки, тот же вид

2 min read

Это LoRA для MiniMax H3 Singularity v1.3 (int8), которая ускоряет обмены в боевых сценах и усиливает удары, сохраняя внешний вид v1.3. Автор — indhic-ai; файл H3_ActionGraft_v1.safetensors весит 155 MB в bf16 и сделан без обучения — рассечением весов двух существующих моделей.

Ключевые особенности

  • Скорость: компонент из файнтюна Singularity v2, извлечённый как rank-16 LoRA; самые быстрые пиксели двигаются на 40–55% быстрее, чем на v1.3.
  • Удары и блоки: text-refiner из Motion Repair V2 (2 из 52 блоков, 16 из 416 тензоров); побочные эффекты других блоков исключены.
  • База: Minimax-h3_Singularity_ref2va_Pruned_v1.3_int8, strength 1.0. Не ставить поверх Singularity v2 или мержей с v2.
  • HyperFlow: при любом LoRA curve fit отключается; чтобы использовать под HyperFlow, нужно merge в checkpoint и строить curve fit для merged file.
  • Промпты: описывать оба конца контакта — удар и эффект на тело.

Пример

Промпт из карточки: “her shield takes the club; she is driven back a step, boots sliding”. Impact-часть LoRA усиливает контакт, который описан в промпте, но не придумывает его сама. Тестировали с Pika H3 flex ref2va LoRA (6 steps, euler/simple), pass 1 608×352 и 1-step DMAD refine 1280×720.

ПодробнееБенчмарки: demon-hunter fight — скорость 28.4 → 42.4; wuxia duel — 39.7 → 51.1; в главном тесте ActionGraft даёт шесть быстрых ударов там, где v1.3 — два.
Время генерации почти не меняется: 158 vs 160 с и 181 vs 189 с.
Ограничения: один seed на тест, смазывание на быстром движении не исправлено, физика лучше, но не идеальна; тестировали только на v1.3 int8 с Pika 6-step LoRA при 720p.
Сделано без обучения: SVD-извлечение rank-16 LoRA и отбор text-refiner тензоров Motion Repair V2.
Видео-примеры
Discuss Telegram 2 12

Will it run?

Runs in 4-bit

Full precision does not fit, a quantised version does: quality barely changes down to 8 bits, more noticeably at 4 bits.

  • BF16weights 62 GB · ready files69 GB
  • FP8 / INT8recommendedweights 31 GB35 GB
  • 4-bitweights 18 GB20 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.

This is a LoRA: counted together with its base model Minimax-h3_Singularity.

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

Run it on Massed Compute

A ready script for the machine we advise above: paste it and the model downloads and starts.

  1. 1. Rent RTX A6000

    $0.43 an hour with the coupon. Pick the Ubuntu image when you create the machine.

    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.

    #!/usr/bin/env bash
    # H3-Singularity-ActionGraft
    set -e
    # NVIDIA driver: CUDA 13 needs 580+. If older or missing, the current nvidia-open is installed (keep yours: KEEP_DRIVER=1)
    DRV=$(nvidia-smi --query-gpu=driver_version --format=csv,noheader 2>/dev/null | head -1 | cut -d. -f1)
    case "$DRV" in ""|*[!0-9]*) DRV=0 ;; esac
    if [ "$DRV" -lt 580 ] && [ -z "${KEEP_DRIVER:-}" ]; then
      . /etc/os-release && \
      curl -fsSLO "https://developer.download.nvidia.com/compute/cuda/repos/ubuntu${VERSION_ID//./}/x86_64/cuda-keyring_1.1-1_all.deb" && \
      sudo dpkg -i cuda-keyring_1.1-1_all.deb && rm -f cuda-keyring_1.1-1_all.deb && \
      sudo apt-get update && \
      sudo DEBIAN_FRONTEND=noninteractive apt-get install -y nvidia-open || \
      { echo "Could not install the NVIDIA driver: see the error above. To keep the current one, run with KEEP_DRIVER=1."; exit 1; }
      echo "NVIDIA driver updated. Reboot the machine (sudo reboot) and run this same script again."
      exit 0
    fi
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    # CUDA 13+ → int8_convrot (Comfy-Org's advice), else fp8_scaled
    CU=$(nvidia-smi | grep -oP 'CUDA Version: \K[0-9]+' || echo 12)
    [ "$CU" -ge 13 ] && TORCH=cu130 || TORCH=cu128
    [ -d ~/ComfyUI ] || git clone https://github.com/comfyanonymous/ComfyUI ~/ComfyUI
    cd ~/ComfyUI
    uv venv --python 3.12 --allow-existing && source .venv/bin/activate
    uv pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/$TORCH
    uv pip install -r requirements.txt "huggingface_hub[cli]"
    # Base files: MiniMax H3, 8-bit
    if [ "$CU" -ge 13 ]; then
      hf download Comfy-Org/MiniMax-H3 text_encoders/qwen3vl_32b_minimax_h3_int8_convrot.safetensors vae/minimax_h3_video_vae_fp16.safetensors vae/minimax_h3_audio_vae_fp32.safetensors --local-dir models
    else
      hf download Comfy-Org/MiniMax-H3 text_encoders/qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors vae/minimax_h3_video_vae_fp16.safetensors vae/minimax_h3_audio_vae_fp32.safetensors --local-dir models
    fi
    # Main model: Minimax-h3_Singularity
    hf download WarmBloodAban/Minimax-h3_Singularity Minimax-h3_Singularity_ref2va_v2_bf16.safetensors --local-dir models/diffusion_models
    # LoRAs (1)
    hf download indhic-ai/H3-Singularity-ActionGraft H3_ActionGraft_v1.safetensors --local-dir models/loras/indhic-ai__H3-Singularity-ActionGraft
    # 16-bit model: ComfyUI holds it in 8 bits (--fp8_e4m3fn-unet), half the memory
    # Start ComfyUI. From your computer: ssh -L 8188:localhost:8188 user@IP, then http://localhost:8188
    python main.py --listen 127.0.0.1 --port 8188 --fp8_e4m3fn-unet
    Add more LoRAs to Minimax-h3_Singularity
  3. 3. Use it

    From your computer: ssh -L 8188:localhost:8188 user@machine-IP, then open http://localhost:8188. Or open that address in the browser on the machine's desktop. The model's workflow is usually in its Hugging Face card.

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.

MiniMax H3 prompts

All prompts

Ready prompts with the resulting video: copy and try.

Model family

Requires the base model

WarmBloodAban/Minimax-h3_Singularity

This model is a LoRA / adapter. Download the base model first, then load this one on top.

  1. MiniMax-H3family root
  2. Minimax-h3_Singularity
  3. H3-Singularity-ActionGraft

New posts about versions of MiniMax-H3 (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 12All on Hugging Face
Merges 3All on Hugging Face

About the model

indhic-ai/H3-Singularity-ActionGraft
Author
indhic-ai
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

122

Ask AI about this model

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

Comments