Self-trained LoRA for the character Thomas The Train from Thomas & Friends
Recommended LoRA weight:~0.8-0.9
Make sure to add "Thomas the Tank" or "Thomas the Train" to prompt.
Works with most anime specialized checkpoints for a more abstract,anime, pastel like art.
Furthermore, semi-realistic checkpoints seem to work great for more show like 3d/CGI style.
Tested with following checkpoints:(Deliberatev2,Meinamixv6,RealisticVisionv13,Stable-Diffusionv1.5)
Feedback is welcome and appreciated, just like advice to improve.
Need help? write a comment!
This was 8 hours of training with over 4000 steps btw. :)
Enjoy.
Description
First Release
Known Issues: Inconsistency,fails to create Thomas' face from time to time.
FAQ
Comments (9)
Thanks I don't know how I lived without this
I know right? 4hrs of training well invested!
Why does this have the NSFW tag?
On second thoughts, I'm sure I don't want to know.
This looks like it was trained on the new CGI thomas series, would you ever think of making a LoRA using the practical models from the early seasons?
Actually haven't thought of doing that. This was more of a meme than being serious. On second thought I believe it would be harder to make a LoRA for the practical models, since there aren't that many great images to train(no pun intended) it on. So yeah, probably won't happen.
@obungaskr wouldnt it be easy to just rip images from a show? alot of still images.
I am Not saying you should, but that would be a feasible aproach.
I guess the one complaint I'd make is how slightly inaccurate it is to the show. The model has a lot of domes and whistle on it. Other than that, it's pretty good.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.