Yujin & Liz
Yujin and Liz are my pets, he is an Akita Inu and she is a Siamese.
I'm having a lot of fun testing the LoRAS that you guys upload here, testing them with Yujin or Liz or both and uploading the images. I would like you to do it too and fell free to share here your own images generated with Yujin and Liz.
I registered an Instagram account to in the future (soon), upload the images there too:
https://www.instagram.com/yujinarts/
Yujin keywords: "yujin", "akita inu", "dog".
Liz keywords: "lizchu", "siamese cat"
They combine well together, but sometimes a dog-cat/akita-siamese mix is generated, or same size (Yujin weighs 40kg and Liz 3.5kg). If you generate only one of them there is no problem
What Yujin likes:
- Cute images.
- Images with jokes, satire, etc. Some mature themes is fine (like gambling, cigars, alcohol, that kind of funny stuff)
- Images having all kinds of adventures in all space and time.
- Images in various artistic styles.
- Images with references to iconic movies or series.
- Hang out with models, singers and actresses (conventional or XXX, only SFW/PG images)
What Liz likes:
- World domination, evil plans, conquer the world.
- Everything Yujin likes too.
What they dislike:
- Disrespectful images.
- Images with their private parts too visible or weird.
Some of the real photos used in the dataset:
Technical information:
Trained here in Civitai, 1024px, increased to 7 epoch, 2 training batch size
Some of the images in the showcase were generated with a previous version of 5 epoch and 4 training batch
190 photos of Yujin, 42 photos of Liz
Description
I trained on Ostris AI-Trainer in the T2V model with the same Flux dataset, but I actually use it for I2V, where it gives even better results and consistency. I can make Yujin appear in any photo by converting it to a video! The HIGH version isn't really necessary.
No Liz in wan 2.2 version.
By mistake I have uploaded a "v2" (previous version that does not work), if you downloaded that version on the first day, I have already corrected it, you must download the "v3", I have uploaded a .zip with both versions V3 high and low too.