CivArchive
    MiniMax H3 Image: Dual Pose Director - Manual VNCCS Skeleton Posing (ComfyUI) - v1.0
    Preview 149703068

    H3 Image · Dual Pose Director — manual posing branch (VNCCS).

    A ComfyUI workflow for MiniMax H3 Image that renders a character in a pose you set by hand: a 3D skeleton is posed manually, the pose conditions H3 Image, and the model generates a high-resolution image in 6 steps while keeping identity from a reference image.

    Branches

    - A (this JSON): manual posing via VNCCS skeleton.

    - B: photo reverse with SAM3D — online experience only in this episode, no distributable JSON.

    Model dependencies (not bundled): minimax_h3_image_pruned_int8_convrot.safetensors, minimax_h3_image_vae_fp16.safetensors, qwen3.5_9b_w4a8_gptq.safetensors, qwen3vl_32b_minimax_h3_int8_convrot.safetensors

    Known limits: identity drifts on extreme poses; 6 steps is fast but you may need a few runs; the build ships without the original embedded capture image (no verifiable licence), placeholders mark where it was.

    Episode: https://youtu.be/Ncg4_CFSaaM

    Chinese original: https://www.bilibili.com/video/BV1nSsr67EHA/

    Manual posing (online): https://www.runninghub.cn/post/2122731136920145921/?inviteCode=rh-v1111

    SAM3D photo reverse (online only): https://www.runninghub.cn/post/2122731130955845634/?inviteCode=rh-v1111

    Description

    H3 Image · Dual Pose Director — manual posing branch (VNCCS).

    A ComfyUI workflow for MiniMax H3 Image that renders a character in a pose you set by hand: a 3D skeleton is posed manually, the pose conditions H3 Image, and the model generates a high-resolution image in 6 steps while keeping identity from a reference image.

    Branches

    - A (this JSON): manual posing via VNCCS skeleton.

    - B: photo reverse with SAM3D — online experience only in this episode, no distributable JSON.

    Model dependencies (not bundled): minimax_h3_image_pruned_int8_convrot.safetensors, minimax_h3_image_vae_fp16.safetensors, qwen3.5_9b_w4a8_gptq.safetensors, qwen3vl_32b_minimax_h3_int8_convrot.safetensors

    Known limits: identity drifts on extreme poses; 6 steps is fast but you may need a few runs; the build ships without the original embedded capture image (no verifiable licence), placeholders mark where it was.

    Episode: https://youtu.be/Ncg4_CFSaaM

    Chinese original: https://www.bilibili.com/video/BV1nSsr67EHA/

    Manual posing (online): https://www.runninghub.cn/post/2122731136920145921/?inviteCode=rh-v1111

    SAM3D photo reverse (online only): https://www.runninghub.cn/post/2122731130955845634/?inviteCode=rh-v1111

    Workflows
    Flux.1 D

    Details

    Downloads
    35
    Platform
    CivitAI
    Platform Status
    Available
    Created
    11/21/2026
    Updated
    11/21/2026
    Deleted
    -

    Files

    H3_Image_3D_____________________.json

    Mirrors