v2.37 — Windows Portable

Windows portable package with dedicated Python, Pillow and drag-and-drop support. No system Python installation required.

Download AI-Visual-Editor-v2.37-Windows-Portable.zip, extract the entire archive and run Start AI Visual Editor.cmd.

v2.36 — Extra Folders

Organise Extras in subfolders with separate folder and item selectors. Leave the folder blank to see files in the Extra root; choose a subfolder to see only its PNG files. Nested subfolders are supported.

v2.35 — Collage Fit

• Optional automatic collage height based on the first image's proportions.
• Optional full-image fit for panels, including images with different proportions.

Both options are off by default.

v2.34 — Text Settings

Edit excluded tags, excluded LoRAs, footer notes and copyright text directly in Settings.

• Four editable fields with automatic saving, shared by Editor and Collage, including ComfyUI filtering.
• All four fields are empty on fresh installations. No personal settings or text files are shipped.
• Existing text files are imported when the corresponding setting is absent; saved empty values stay empty.
• Footer and copyright placeholders remain supported.
• Independent signature opacity sliders for Editor and Collage, applied to previews and exports.

v2.33 — Collage Grid

Create collage grids with configurable columns and rows, such as 2 × 2.

• Images fill cells from left to right, then top to bottom.
• Additional rows are added automatically when needed; empty cells use the gap colour.
• Grids also work as a starting arrangement for free layouts.
• Each image retains its independent scale control.
• Fixed numeric scale entry: typing a value no longer inserts the slider's minimum value while editing. Enter or leaving the field applies the limits.
• Updated the README with grid instructions.

Both the grid and numeric entry fix were tested successfully by the user in the application.

v2.32 — Collage Image Scale

Each collage image can now be scaled independently inside its panel.

• Scale slider and numeric entry from 10% to 300%, with a reset to 100%.
• Works with automatic grids and free layouts, preserving image proportions.
• Preview updates immediately and exports use the same scale.
• Exposed space uses the selected gap colour; crop positioning remains adjustable.
• README updated with the new controls.

Validated locally by the user in the application.

v2.31 — Extra Targets

What's new in v2.31

Every extra element now chooses where it goes. Print on used to be a single setting for all of them: either everything went into the thumbnail frame, or everything went on the full picture. Now each badge carries its own destination, the way text layers already did — so a discount badge can sit inside the thumbnail while a watermark goes on the full image.

The selector applies to the selected element; with nothing selected it sets the default for the next one you add. Each row in Layers Control shows where that element is headed (🔲 frame, 🖼 image).

Worth knowing: moving the selector no longer retargets the extras you have already placed, only the selected one.

v2.30 — ComfyUI

What's new in v2.30

ComfyUI support. ComfyUI stores its whole workflow in the image instead of a prompt block, so these images used to produce a text file saying No metadata found. The tool now saves the workflow as a .json next to your exports — drag it into ComfyUI and the graph opens as it was.

Your filters apply to it. ignore_loras.txt and ignore_tags.txt clean the JSON too, so it never reveals what the text file would have hidden:

• LoRAs are matched by file name, whatever folder they are in. Power Lora Loader entries are removed entirely; in base loaders the path is blanked in place.
• Excluded tags are removed from every prompt in the workflow, <lora:…> calls included, while your formatting stays as you wrote it.

The exported image is still stripped of metadata unless you enabled Metadata in Settings. In the Collage, ComfyUI images get their own numbered JSON.

v2.29 — Shaped Frames

What's new in v2.29

Shaped Frame Mode — formerly Circle Texture Mode. The name no longer fits: the mode handles any shape, not just circles.

Fixes

• Choosing a different overlay now updates the preview immediately, as normal frames do.
• Exporting the same picture with several overlays no longer overwrites the previous file: the overlay's full name goes into the file name (picture_cerchio_film-1_account.png).

v2.28 — Parameters

What's new in v2.28

You choose which parameters end up in the text file. Extensions bury the useful information under their own: Ultimate SD Upscale adds five fields, ADetailer around fifty. A new panel in Settings keeps only the parameters you pick, in the order the webui writes them — and skips the ones missing from a given image instead of printing them empty. Excluding something is simply leaving it off the list.

It comes ready to use, with the essentials already selected. A dropdown offers a catalogue of ~90 known parameters grouped by origin, Read from image lists the ones a loaded image actually carries, and wildcards let a single entry cover a whole family — necessary for extensions that number their units, like ADetailer model 2nd or ControlNet.

The model, its hash, the LoRAs and the refiner are untouched by the filter: they keep their own Models: section with their links.

Note: the separate Include RNG in parameters checkbox is gone. RNG is now one entry in the list like any other, and it is enabled by default.