LoRA for Gwendolyn Tennyson from Ben 10.
*Refer to version notes for EACH LoRA and how to use them. It is different for each of them and the information below only applies to the first version of this model
I trained it with 74 pics, which was a mix of mainly fan art and some screencaps. I trained it at 10 epochs with a dim/alpha of 16/8.
Leans to a semi-real look mostly, but somewhat more cartoonish at higher LoRA weights. Let me know how it goes for you. Read stuff below on guidelines and known issues :)
Triggers:
Input "Gwendolyn Tennyson, 1girl" after the LoRA near the start of the prompt.
Input "orange hair, green eyes" for more consistency if needed, but I didn't find it necessary.
Input "long sleeves, white pants" for classic clothes. NB: can place in negative prompts if the clothes are stubbornly appearing.
Notes:
My experience has shown scale 0.7 ~ 1 works best; my best results were 0.7 for semi-realistic, but overall favourite was 1.
My prompt order shown in the sample pictures worked best for me.
Works on different models though weights may need to change for good results.
I used Anything4.5 model and Midnight Maple model. Midnight Maple gave best results for this LoRA for me.
Refer to my pictures for examples and what upscaler I used, keep in mind I have used other textual inversions to help with poses and negatives.
I think this is personally very flexible, at least in terms of changing poses, clothes etc. Other LoRA's and/or Models seem to bring out the more cartoon appearance so experiment for styles.
I don't think people read entire descriptions like this, but congrats if you do
Description
Lora Update for Gwendolyn Tennyson (Ben 10)
Hi all☺️ Been gone for quite some time but finally back. Decided to update my Gwen LoRA to the XL version (specifically Pony) since I've had numerous requests so far.
Essentially, it's been trained on the same dataset with tweaks made to settings so that it turns out well for Pony and similar Mixes. Note that this does not cover the Lucky Girl version, but if there's demand then I'll make one for that too.
This LoRA will only need one tag as usual and remains flexible. Please play around with the LoRA weights, but a weight of "1" worked fine for me. Read stuff below on guidelines and known issues :)
Triggers:
Gwendolyn_TennysonIf you want to get her classic outfit, then the following tags may help:
Blue shirt, long sleeves, long white pants Notes:
My experience has shown scale 0.9 ~ 1 works best; my best results were 1.
My prompt order shown in the sample pictures worked best for me.
Works on different models though weights may need to change for good results.
Refer to my pictures for examples and what model and upscaler I used.
I honestly did not have much time to play around with this one and is my first XL Lora, so seeing your works will help me refine it if need be. Don't forget to comment too😊.
FAQ
Comments (18)
Glad to see Pony version. Version name (above pictures) is labeled XL, but the base model is listed as Pony.
Thanks, I've changed it to pony to avoid confusion. I think it's technically still an XL model since Pony was based off that (could be wrong). In any case thanks 👍
@reevee Pony is SDXL. It's just such a finely tuned version of SDXL1.0 it basically got its own label since LoRas and stuff can have a drastically different outcome of generation when combined with something non-pony. There is a SD1.5 version of pony as well, but anything labeled "pony" on civit is pretty much talking about the SDXL version.
Does it also support the lucky girl costume?
Hi :)
No this does not. It's specifically for the V1 version of my lora hence why I named this one v1 (Pony) but I'll look in to adding the v2 (Pony) which will have the costume at some point later. But of course if theres a big demand for the lucky girl one then I'll do it sooner
Nice 🙂 👍
Would you be up for making a lora for the OS show style in general?
Hi :) would you mind elaborating? Did you mean OG style or does OS stand for something else😅
@reevee Oh right yeah the original show like:
https://civitai.com/models/84097
I'm surprised that there isn't a PONY for it yet.
Sorry, didn't seem to get the notification for your reply I think hence the late reply. I will definitely keep it in mind, I think if enough people want it then I'll make it. You can commision me if you really want it, but I don't expect you to since it's not like everyone can afford to
How do you make it such a beautiful cartoon style while seeming it has depth, all of mine are either too realistic or picasso and bad 2d
Hi thanks for the comment :)
There's quite a few things that contribute to it and it's mainly my own theory about how I do my stuff differently. First I train on a model that can produce that style (Pony in general can do that so this step is not an issue). Of course the images generated by me would also need to use a model good for that style and appropriate prompts, so if you want something close to my style then copy my prompt style. (notice my stuff with this lora looks different than others who use it. A lot of the time Loras can be forced to look a certain way even if they're a bit strict with style). Additionally, in order for a lora to be flexible with style I actually train with a mix of screenshot and fan art (at the start of most of my lora descriptions I mention how many images I use and how many were fan art. I almost always aim for 50 to 75 pictures but sometimes go over 100) I try to use a good mix with various situations, clothes, facial expressions, actions and make sure the fan art is varied but all fan art that seems pleasant to look at and still quite close to the original cartoon style (note the time consuming part about how I make my loras is that I manually go through and pick each one and crop them, and even edit out unwanted people or things. I try to keep the images close to square if I can and remove other people or animals from the image if possible). Also, the settings of the LoRA are important: I use hollowstrawberry Google colab lora trainer and for the most part use a lot of the default settings though I usually change the learning rate to accommodate for longer training and the many pictures I use
I'm still amazed at how good this lora looks, ironically not resembling Gwen in the cartoons. I'm more of an anime kind of guy, and I don't really care much for Ben10, let alone gwen.
Can someone help me set up AI on my PC?
Hello, i recomande you this chanel https://www.youtube.com/@Aitrepreneur .
And if you have an issue, use chatgpt for helping you to resolve probleme.
I'd suggest trying Stability Matrix - it's a sort of hub that will do the hard parts of setting things up for you and help you to keep the models you download organized.
It also has a built in tab called Inference, which is basically a very simplified UI for ComfyUI with optional autocomplete for tags. ComfyUI is faster and produces better results (in my opinion) than the alternatives, but it can get complicated - so Stability Matrix is nice because it lets you ignore the complicated parts while still getting good quality results.
You can download it from their site here: https://lykos.ai/
They also have a Discord server with a support channel that's quite active if you have any issues.
If you have WIndows/Mac It's as easy just use the ComfyUI installer. Go to their official website https://github.com/comfyanonymous/ComfyUI?tab=readme-ov-file ..
There should be a desktop application link. It's actually the very first one there. You'll see it when you scroll down the page a little bit.
If you're on linux there should be instructions on how to install it on the github page.
is this mfker for real? just click the installer
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.