So basically, it's a text improvement LoRA, but it also acts as an aesthetics enhancement LoRA (which was unintended, but still welcome). Beta version.<h2 id="how-to-use:">How to use:</h2><ul><li>There are three main trigger words: "speech bubble", "text", and "snapchat".</li><li>While the first two are self-explanatory, "snapchat" adds a Snapchat-like semi-transparent bar with text on it.</li><li>The basic pattern is: {trigger word} "words words words".</li><li>You can also use [n] in some cases to create a speech bubble or a string of text (Unstable!).</li><li>Just look at the prompts in the example images :3</li><li>You can also use this lora without any trigger words, just for an aesthetics enhancements.</li></ul><h2 id="train-info:">Train info:</h2><ul><li>15 epochs, 2925 steps</li><li>Dataset: ~400 imgs, manually captioned text, WD3 Large for tags</li><li>Batch size 2, gradient acc 4, keep tags 5, shuffle the rest, no dropout, TE was NOT trained.</li><li>Resolutions = [768, 1024, 1280, 1536]</li><li>Trained on WSL2, RTX 5060 Ti 16gb for ~16 hrs.</li></ul>I want to continue working on this LoRA; however, captioning the text by hand is a huge pain in the ass.

Textify [Beta]

Doubled the dataset's size, switched to LoHA, tried to train LLMAdapter alongside. Trained on Base-v1. Stronger aesthetics? Don't expect much.

Files with SHA256: 66166107fa45a9d6efde60654eb292c819cca877e02f0e32a4392aa39d9e974f