MOSS BBOX Prompter for ComfyUI

A visual bounding-box editor for MOSS-Image, adapted from Krea2 BBOX Prompter Suite.

Place subjects and text on a canvas, describe each region, and export the layout for your MOSS-Image workflow. Use Character, Object, and Text types to organize people, objects, and lettering.

Now standalone: Krea2 BBOX Prompter Suite does not need to be installed separately. The required UI components and helpers are included.

The interface was developed and tested with ComfyUI’s classic node interface. Nodes 2.0 is not required.

• Visual BBOX positioning and resizing.

• Separate Canvas and Prompter nodes.

• Character, Object, and Text region types.

• Canvas presets, camera settings, and custom image dimensions.

• Two example workflows: Standard and DMD, with WebP output.

• Existing MOSS BBOX workflows remain compatible.

Models

• MOSS_LUC — My Anime Model

• MOSS-Image SFT — ComfyUI INT8

Text Encoder / VAE / LoRA

• Text Encoder — Qwen3.5-9B INT8

• VAE — FLUX.2 Small Decoder

• DMD Acceleration LoRA — MOSS-Image SFT 8-Step

Changes v1.0.7

• Lock/unlock selected BBOXes beside X/Y/W/H. Locked boxes remain selectable and reject movement, resize, numeric edits and deletion. Saved with workflows; Undo/Redo supported.

• Connect Load Image IMAGE to Canvas reference_image for a reference underlay without running generation. Default OFF, opacity 50%; loading or toggling display keeps Canvas dimensions unchanged.

• Match image size sits beside the display toggle. It explicitly fits dimensions to a multiple of 16 and scales boxes, with Undo/Redo. The reference is a layout aid and does not condition generation.

• Japanese/English help and README include wiring instructions and the supplied usage screenshot.

v1.0.4

Added Japanese/English tooltips, fixed Preview disappearing when zooming out, and stabilized the coordinate controls. Updated Standard and DMD workflows now include a connected BBOX Preview.

Download both workflows from the GitHub Release. Classic ComfyUI UI is supported—no Nodes 2.0 or separate Krea2 installation required!

V1.0.0

includes the Canvas editing improvements introduced in v0.1.4:

• Undo / Redo: restore drawing, moving, resizing, deleting and resetting BBOX layouts, with up to 60 edits per Canvas during the current editing session.

• Keyboard shortcuts: Ctrl/Cmd+Z to undo; Ctrl+Y or Ctrl/Cmd+Shift+Z to redo while the Canvas or its history buttons have focus.

• Overlapping BBOX selection: Alt+click to cycle through overlapping boxes. The selected box has a white outline and can then be dragged normally.

• Canvas shortcuts avoid accidentally triggering ComfyUI’s workflow-level Undo.

Installation

Open a terminal in your ComfyUI folder:

cd custom_nodes
git clone https://github.com/ukr8b3g-cmyk/MossImage-BBOX-Prompter.git

Restart ComfyUI, then refresh the browser. If the old interface remains visible, press Ctrl+F5.

Install the MOSS-Image model files and any additional nodes required by your chosen workflow separately.

Updating

From your ComfyUI folder:

cd custom_nodes/MossImage-BBOX-Prompter
git pull

Restart ComfyUI and refresh the browser after updating.

Repository and workflows

Visit the repository for documentation, example workflows, updates, and issue reporting:

MossImage-BBOX-Prompter on GitHub

Download the example workflows

Based on the MIT-licensed Krea2 BBOX Prompter Suite, with the interface and internal processing adapted for MOSS-Image. Distributed under the MIT License.