Anima Edit | Personification

Given a source image, create a personification of it.

Usage

ComfyUI

The workflow is included in the preview images, you can download the images and drag them into ComfyUI.

The ComfyUI-Cosmos-Reference custom node is required at minimum to use the LoRA with Anima.

Forge Neo

Forge Neo also supports cosmos reference, so you can use this LoRA in the img2img tab. (Make sure to enable it in Settings->Stable Diffusion)

Prompting

Use natural language with the clearly worded edit instruction including the word Personify , and tags with the personification trigger.

The preview images all use the ideal prompt structure, which is something like:

Create a personification of a lapis lazuli gemstone featuring Hoshimachi Suisei with blue hair and a blue dress with gold and silver trim.
personification, masterpiece, best quality, hoshimachi suisei, 1girl, solo, full body, blue hair, very long hair, blue dress, kintsugi, starry sky, night sky, blurry background

Explanation of the workflow

This workflow uses this LoRA to recreate personification artworks that include an embedded reference image in the work itself.

The source image is a real image that is loaded and resized to a 1024^2 pixel SDXL resolution (832x1216) using the custom JPS-Nodes, and the ComfyUI KJ-Nodes - Resize Image v2 node.

*The resizing is important to keep the source image in a normal range that matches both Anima's training and this LoRA's at 1024^2 pixel resolutions with sides divisible by 16.

The Anima Model and this LoRA are all loaded normally, the PC Schedule Prompt custom node is optionally used mix artists in the prompt with the [artist1|artist2|artistN...] format.

The ComfyUI-Cosmos-Reference custom node is used to apply the Cosmos latent reference of the source image to the generation, which is required for this LoRA to function.

In the "Save" group, the Draw Text, Resize and Image Composite Masked nodes are finally used to write the "Reference Image" text above the source image, and compose it over the generated image in the top left corner.

Compatibility

The Anima base model is not natively trained for edit tasks, stacking this with other (non-edit) LoRA or reducing its weight can degrade the outputs quickly.

Notes

This LoRA is just a bit of fun, it is very creative and not very suited to precise edit tasks.

This is my second attempt at training Anima with edit datasets, first one being here: YAAE | Yet Another Anima Edit.

I found interesting how the same training pipeline produces very different outcomes for precise edit tasks compared with this very vague task of personifying arbitrary things.

This first LoRA might not make any change if the prompt was not targeted/represented well enough in the training dataset, but this one will completely change the image regardless of prompt.