This LoRa try to mimic the simple illustration style from kids book.
!!!!!!!!! PLEASE DON'T POST LEWD IMAGES IN GALLERY, THIS IS A LORA FOR KIDS ILLUSTRATION !!!!!!!!!
Training LoRa necessitate a lot of coffee. If you think that what I do is useful and want to help me keep doing this you can buy me a coffe here: https://ko-fi.com/clumsy_trainer
Description
Re-trained from scracth. Results may be different 'cause I lost the first dataset. The shading and overall prompt coherence should be better, tho
FAQ
Comments (51)
Amazing Work
Share your negatives in a single zipped file...
Do I needd "kid" keyword? thanks.
sometimes it helps but it's not mandatory, I didn't use that word in captioning for every image, just for some
I'think your change very good
Excellent work on the update. Super happy to see this getting the attention it deserves. Looking forward to seeing what you make next!
thanks! I'm so glad that people liked it
i really love this model, if you looking to train a model similar to this, look into i hate fairyland style, it really bright and pretty.
Please teach about the painting style training is how to do it, if you have time hope to come out with a video tutorial, thank you!
nice one
really useful for commercial this kind of design style
I can't get this to run. How do I use it? I tried using ReV Animated and v1-5-pruned-emaonly but all I get is noise. I'm using easy diffusion, might this be a problem?
tbh, I dunno 'cause I've never used Easy Diffusion, does it supports LoRa networks?
@Clumsy_Trainer I think it does. Other LoRas work quite well :/ Though it can only have a strength between 0 - 1
@Clumsy_Trainer I tried one of your other models (2000s Comics), this works fine. It seems to only be an issue with Kids Illustration. Do I have to apply specific settings?
@nrczbol864564 I can't rememeber if in one of the merges I made to obtain the version posted here there's a lycoris or LoCon model, This is the only thing that comes into my mind that could give problems
When I go strength 1, it gives only a group of people.
Is there a specific base-model you recommend using this LORA with? What about weight?
hi, all images were generated with realisticVision V2 but you can use any model based on 1.5, some of them tends to anime tho. For the weight, it was trained to be used at 1.
@Clumsy_Trainer Thanks, this is useful!
Why is this the only Lora that is working for me? I tried blindbox and Studio Ghibli Style LoRA but they just dont work. Help please.
sorry, I have no idea, it's strange tho. One thing, maybe it's a LyCoris or LoHa, in that case you'll need an additional extension
I know that the most recent update to automatic1111 broke most of my loras and embeddings, so try downgrading a bit.
try trigger words
I'm sorry, i'm pretty new at this, but how to trigger this style? I already put the lora in lora folder. Any trigger word to put in the prompt?
If you are using sd webui then just click on the icon below the generate button. That will activate a submenu where Lora is
this is a great work!!
thanks!
Hope to continue to update this type of illustration lora, this lora style is really good
I love v2 style, too bad it alway generate tree, forest in corner
I keep getting 2 characters not purposely , anyone have same issue?
what model are you using as a base? try the same prompt without the LoRa and see if it gives you more than one char. If it does, it's your checkpoint. Another thing you can do is use the "1boy" or similar token in your prompt. Hope it helps, let me know
Hi, Thank you so much for your reply, I use model: RealisticVisionV2.0 which I think this are the one that is used very often in the gallery,
here's my prompt:
1kid, short straight hair with bangs, 4 yo, cute face ,cute expression, wearing orange T-shirt and sling overall jeans ,outdoors, background forest, toon, pixar style, 3d, cartoon, detailed face, full body <lora:COOLKIDS_MERGE_V2.5:1>
and negative prompt:
easynegative, ngtvH, ngtv, ngtvB, bad composition, glitches, (bad quality), deformed, distorted, bad drawing, ugly, blured, realistic, symmetric, bad hands, extra child, cross-eye,
The extra child is still appears.
I tried turned off the LORA I did not get the extra child
I'm having the same problem, it's drawing multiple character... really annoying.
What VAE is used?What I make is always a lack of sunlight.
no VAE used, try to incorporate "sunlight" in your prompt if you didn't already. Also, I need to know what model you are using for further directions
Do you have any plans on training the model based on SD 2.1? If you feel like it is too much of an hassle, I am open to training it for you and sending to you so you can upload it.
Asking again as SDXL 1.0 has been released. I really like this model, but would like to create one for SDXL. I have a really good system, and would love to train the model for you if you could provide the dataset you used for it. I do not need recognition, just need it for personal use.
realisticVisionV20_v20，Where should this checkpoint be found?
it's in the Realistic Vision page here on CivitAi
what is the link of the checkpoint?
Me and my friends love this cool kids lora very much, we really with this lora supply SDXL one day! :D
Thanks for your kind words, I really appreciate that. It's not easy for me to train on SDXL since I have a GPU with only 8gb of vram. If you want to help, just rate my LoRa or post an image made with it here on civitai, this would help me get buzz to train a SDXL model.
Great LoRa for Warpfusion
Really hard to only get 1 character (i.e : One Elephant) with this LoRa. I'm using the same model as you but i keep getting 2 character even if the prompt says "Alone" or "1 elephant"...
Cant do the empty landscape with this LoRa, without any character and etc, maybe someone has something like solution for this problem?
that's strange, it may depend on the underlying model you are using. I just generated a temple in a desert landscape without any problem (and no character whatsoever) using Dreamshaper 8
Would they flux Lora for this model
I'm about to start working on it, hope I'll be able to share a V1 soon
Thank you amazing gentleman for sharing this with the world.
Thank you for this! I'm trying to make a kids story for my daughter and it's really annoying how hard it is to get kids stuff and how creepy it feels because of all the fk wits.
Is there any way to prompt for one specific consistent style?
Also if you have three datasets used that would be cool. Would love to try a lora with ome style.
Details
Files
COOLKIDS_MERGE_V2.5.safetensors
Mirrors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
Kids.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
artist-kidsillustration.safetensors
COOLKIDS_MERGE_V2.5.safetensors
model.safetensors
COOLKIDS_MERGE_V2.5.safetensors
5abe027b-fc13-4bc4-a23a-5f3e66a019ae.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
pytorch_lora_weights.bin
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
COOLKIDS_MERGE_V2.5.safetensors
461A1DC302A1BD5E25FCE75644E50EC5.safetensors