Update June 28th added Custom A version and pruned version to other models.
Making Children's Books, Comics, Graphic Novels is rewarding, but very heavy on the artwork. This model will help you tell any of your stories visually. Let your imagination go wild. If you need custom characters that are consistent throughout a story, you can hire me to make one for you. Contact me in one of the links under my profile.
Follow me to make sure you see new styles, poses and Nobodys when I post them. Things move fast on this site, it's easy to miss.
PROMPT RESPONSIBLY
How to use:
To get similar results as in the preview images. I recommend using the vae-ft-mse-840000-ema-pruned and clip skip 1. Also check out this Detailed tutorial on how I get the results in the preview images. The model is capable of different illustration style other than the ones shown in the preview images. Experiment with styles or use illustration style LoRAs to dial in the specific look you desire.
Three models included:
V1 Toon-Anime and V1 Semi-Real. You won't be able to make photorealistic images with these model but that's not the point. Create colorful, fun, and imaginative images that excite a child's wonder. V1 3D is specifically designed to give a 3D movie look but not similar to studios that have expensive lawyers.
Do you have requests? I've been putting in many more hours lately with this. That's my problem, not yours. But if you'd like to tip me, buy me a beer. Beer encourages me to ignore work and make AI models instead. Tip and make a request. I'll give it a shot if I can. Here at Ko-Fi
Description
Stable Diffusion 1.5 (512 x 512)
Children's Stories V1 3D
VAE recommended vae-ft-mse-840000-ema-pruned.ckpt
FAQ
Comments (32)
My Gawd Zovya, don't you ever sleep? Yet another fine resource from you. You're one of the most prolific people on Civitai and we're all the better for it. Keep up the great work, I truly appreciate you, you have helped my AI art process tremendously.
Thank you. I have been spending too much time at this lately. Comments like these make it worth it.
these are really great models, , any plans for a realistic one to complete the range
Thanks. No I probably won't since it would really be duplicating my A-Zovya Photoreal model which is capable of this already.
I get not enough precision errors for the 3D model, the others work fine. any idea?
the 3d version is pruned. should be the opposite problem. try using or not using --no-half in your startup script
@Zovya unfortunately adding those lines have no impact. it doesn't work I get: NansException: A tensor with all NaNs was produced in Unet
@dilectiogames ok, try either restarting auto1111 completely or re-download the checkpoint. probably a hiccup somewhere there
@Zovya re-downloaded the model and now works. it looks like civitai now will deliver corrupt files
Relly Nice Model!
Could you tell me what models you've used to create this mix (perhaps a full merge receipt, if you still have it)?
From what I've tried so far:
this model works REALLY well with both, Booru- and CLIP-Type promts while also playing nice with Loras trained on different models. A realistic version of this model would be nice, since the semi-realistic version deforms proportions too much.
(I've also tried your realistic model, but the compatibility with loras and promt-response was much worse than of this model)
Really nice! Is it possible to make LORA versions of these styles? I think people will love them! Thanks a lot!
thanks. good idea. I'll get them started.
Is Clutter-Home a personal TI? I see it in a lot of your prompts and it seems extremely useful, but if you're keeping it as a personal tool for contract work, that's totally understandable.
This is by far my favorite model, the style is amazing.
But for some reason I'm getting this weird problem where eyes keep getting duplicated on top of each other. So you get amazing art, but you can see 4 eyes, 2 normal ones and then 2 more on top of them.
wow, that is weird. that can happen if you generate the initial image too large. what size are you trying to do?
Also this tutorial can help a little if that's the case. Detailed tutorial on how I get the results in the preview images.
@Zovya I always go 512x1024 or maybe 612x1024. I never have that problem on other models, but yours for some reason does that. I'll check that tutorial. Thanks.
I really like what you do. I'll be doing my first book for my daughter with your work. Thanks.
If you need additional GPU power, let me know. My 3090 is idle most of the day and I don't have time to build more datasets right now. I already have dreamtrainer2 setup so you just have to provide the dataset and I can finetune the checkpoint of your choice for you. (You can even provide the training settings if you prefer but I got very good results with what I'm using.) I usually do a first run at 768 and a second at 640 res. If your dataset allows it, I can do a batch size of up to 6 without gradient checkpointing. I got a second GPU for display so I can use the full 24,500mb of VRAM. I built this computer specifically for training ;)
Thanks, I'm excited you're using it to make a childrens book!!
@Zovya I'm going to finetune her into the checkpoint and make her the hero of the story. She will be so happy about it! (I might finetune some cartoon zombies as well)
One of the best, most likely the best checkpoint that i can think of. Absouletly fantastic.
wow. thanks
Hi @Zovya , I've a question. I'm trying the toon version of your checkpoint (the first). Did you add support to quicksand or mud pit? This aspect is part of any adventure novel in general. When I write a prompt for this, it returns a puddle of clear water...
hmmm. maybe try to describe it to the model. "brown muddy water"? prompting is a skill all in itself
Can you share your Upscaler config that you use in you checkpoint?
I use this for all image generation Detailed tutorial on how I get the results in the preview images.
Any chance of making pruned versions of the toon-anime and semi-real versions? I'm sure all the <8gb folks would appreciate it! Love your work btw!