It's another SD1.5 synthetic dataset time!This one was a hasty job. Did some dynamic prompts and called it a day. No captions, some images are not upscaled and content is shallow. But that's still plenty for Anima.I'm supplying a lora based on this dataset too (that's probably most, anyone would even need anyway).I purged the metadata - don't look for any. Just because I might still work with similar style later. But, if need be, you can just use this lora to generate more data very easily with Anima.It's an anime style with a small dash of toonish look. Similar to the 'Qt Lad' lora I uploaded a day ago, but with more variance and stronger style. 250+ images. Trains very quickly. 5 epochs I did, might even be overdoing it.Enjoy!As usual, no credit necessary. Use this however you want.(Try it with 'Artistic Chad Style' Lora at 0.5 - adds some manliness and I find it looks somewhat cooler)(Added a 2 epoch version for comparison - style is less consistent, but output is less stiff, due to dataset shallowness)

DIY ReHusbando Dataset

toon-style-lora.safetensors

ReHusbando-Dataset.zip

Toon Style Lora

2 Epoch Version

<h3 id="some-reorganizing">Some reorganizing</h3>This will now be the place with all the synthetic datasets and associated loras.<hr />'chartact' is new - male character with a mild hint of Akihiko Yoshida's exaggeration (FF Tactics).Dataset is untagged. After a bunch of training, it almost seems like tagging barely matters for Anima. At least with such narrow datasets as these.Trigger is unnecessary either. I add it "just in case".<hr />'toon' has two datasets, one was made in SD1.5 and the other later in Anima from lora trained on first dataset.I did not retrain SD1.5 dataset on Preview3. At least not yet. Busy with other things and not sure it's even necessary.

Files with SHA256: 99abebd88d8a0f3ff64f45c40b2519f6a4c19ee3691d545d36dc922047642f08