CivArchive
    MiniMax H3 Image: 6-Step 16MP Blur Reconstruction (ComfyUI) - v1.0
    NSFW
    Preview 149674689

    MiniMax H3 Image · 6-step 16-megapixel reconstruction.

    A ComfyUI workflow that rebuilds a blurry or low-resolution photo into a sharp ~16MP result in 6 sampling steps: the H3 Image edit chain is conditioned on the source image, then re-sampled and decoded at high resolution.

    What it is for

    - Restoring blurry photos / screenshots / compressed stills

    - One-click upscale-to-rebuild rather than plain interpolation

    - Batch frames pulled from video

    How to use

    1. Load the JSON into ComfyUI (H3 Image nodes installed).

    2. Feed the source image; keep the default 6-step schedule unless you need a different detail level.

    3. Run; the SaveImage node writes the 16MP result.

    Known limits

    - It reconstructs, it does not invent detail that is not implied by the input; heavy compression artifacts may be re-interpreted.

    - Needs the H3 Image model set (four weights listed in the README).

    - Not a video model: process frames individually.

    Episode: https://youtu.be/gnz22jMIK0Q

    Chinese original: https://www.bilibili.com/video/BV1bSs76FEVm/

    Workflow experience page: https://www.runninghub.ai/zh-cn/post/2121694628231987202/?inviteCode=rh-v1111

    Description

    MiniMax H3 Image · 6-step 16-megapixel reconstruction.

    A ComfyUI workflow that rebuilds a blurry or low-resolution photo into a sharp ~16MP result in 6 sampling steps: the H3 Image edit chain is conditioned on the source image, then re-sampled and decoded at high resolution.

    What it is for

    - Restoring blurry photos / screenshots / compressed stills

    - One-click upscale-to-rebuild rather than plain interpolation

    - Batch frames pulled from video

    How to use

    1. Load the JSON into ComfyUI (H3 Image nodes installed).

    2. Feed the source image; keep the default 6-step schedule unless you need a different detail level.

    3. Run; the SaveImage node writes the 16MP result.

    Known limits

    - It reconstructs, it does not invent detail that is not implied by the input; heavy compression artifacts may be re-interpreted.

    - Needs the H3 Image model set (four weights listed in the README).

    - Not a video model: process frames individually.

    Episode: https://youtu.be/gnz22jMIK0Q

    Chinese original: https://www.bilibili.com/video/BV1bSs76FEVm/

    Workflow experience page: https://www.runninghub.ai/zh-cn/post/2121694628231987202/?inviteCode=rh-v1111

    Workflows
    Flux.1 D

    Details

    Downloads
    2
    Platform
    CivitAI
    Platform Status
    Available
    Created
    11/20/2026
    Updated
    11/20/2026
    Deleted
    -