Generate pictures with Arthemy
Searching for an AI characters generator?
Arthemy, my AI image generator now has all of my models loaded and ready to be tried on your browser - no GPU required
----> https://arthemy.ai
✦ Comics Arthemy - Quick Start ✦
1_____ I highly suggest to use this embeddings:
verybadimagenegative_v1.3 https://civarchive.com/models/11772/verybadimagenegativehttps://civarchive.com/models/11772/verybadimagenegative
2_____Size: Around 640 x 768
3_____Prompt: Start from the prompt present here and tune the results based on your needs:
Quick Prompt:
[How],[what/who],[where],[other] exaple: (ultradetailed fantasy illustration:1.3), elf witch woman with a large hat and purple dress, on a snow-covered mountain with blowing wind, complex lighting and strong sense of movementQuick Negative Prompt:
verybadimagenegative_v1.3, (worst quality:1.4),(low quality:1.4),(normal quality:1.3),lowres,watermark, title, (jpeg-artifacts:1.3)
4_____Upscale: to get crisp results like the one shown in the pictures use HiResFix and Lanczos. I suggest to use this step only on the preferred pictures, since it's quite heavy on the performance side.
I hope you like my merge and I can't wait to see what you all are going to create with it!
For future update, about the AI Editor I'm building with an italian team and my models you can check:
https://www.instagram.com/arthemy_ai/
____________________________________
Merged / subtracted / inspired by Models:
DarkArtsStyle - https://civarchive.com/models/109?modelVersionId=124
DiffusionBrushEverything - https://civarchive.com/models/46294?modelVersionId=50908
ForgottenMixtheNovelist - https://civarchive.com/models/81644?modelVersionId=86640
Icantbelieveits - https://civarchive.com/models/81644?modelVersionId=86640
icomix - https://civarchive.com/models/16164/icomix
mcbsMachinecodesComics - https://civarchive.com/models/71404?modelVersionId=76148
OccidentalMix V2.0 - https://civarchive.com/models/63920?modelVersionId=68515
vividwatercolors - https://civarchive.com/models/4998/vivid-watercolors
westerncartoontypeA - https://civarchive.com/models/62060?modelVersionId=66582
Babes - https://civarchive.com/models/2220/babes
Diffusion Brush Everything - https://civarchive.com/models/46294/diffusion-brush-everything-sfw-nsfw-all-purpose-checkpoint-nuclear-diffusion-anime-hybrid
Cute Cartoon - https://civarchive.com/models/85547/cute-cartoon-illustration
Western animation diffusion - https://civarchive.com/models/86546?modelVersionId=92044
IdentityCrisis - https://civarchive.com/models/89075?modelVersionId=94798
Vivid Watercolors - https://civarchive.com/models/4998/vivid-watercolors
Exquisite details极致华彩 - https://civarchive.com/models/118495/exquisite-details
SXZ Luma - https://civarchive.com/models/25831/sxz-luma
+ some of my old mix
Some of these models have been used for subtraction, but they all have been useful in order to create this model.
If you like Comics Vision, remember to check these other models and consider to support their creators!
Description
This is pretty much Arthemy Comics v7.0, but with Baked VAE in ordedr to solve the graysh and desaturated outputs.
FAQ
Comments (19)
Arthemy! Can you make like "full-sized-model" in fp32 and about 5GB+ weight?? Will it increase quality or divergence somehow??
I don't know anything about making models. but I think arthemyComics is probably the best v1.5 model ever. Maybe it can be better?
Can it be better? Sure! But it truly depends on what you want it to generate.
I've made thousands iterations before reaching this version and my main objective was to create a very flexible model with a lot of different styles, able to reach consistent results with *most* of them.
In my experience, to improve this model further, some choices needs to be made - for example, abandoning some flexibility in order to gain more quality in one of the specific style.
You can imagine the models as limited boxes completely full of informations and, since the box is always full, you have to leave something behind every time you want to add something new.
It you want some help we can have a chat and we can think together how to improve this model, I'm more than happy to help! 😌
@Arthemy I thought about general increase in technical quality, like ability to produce more details to whatever image in whatever style is made. Making model "not pruned" won't help with that??
Ps: Full-sized models and pruned fp16 models bring to the same exact results. The extra weight is informations that doesn't directly or indirectly affect the generations - somehow.
@Arthemy shock... why than I'm allways using full ones when available...
@Arthemy well then, in matters of content-diversity present model is exactly what is needed
@oddKoodoo do you want more technical quality and details? Sure! But what are you willing to sacrifice for that? Less variety of styles or less variety of character?
It's always like that, especially with lightweight models like SD1.5.
You can try to reach your specific style with Loras, they seems the kind of "aesthetic/technical" tuner that you might need to start and experiment with.
@oddKoodoo May I suggest to try ComfyUI and Arthemy XL in order to see if that increase in size (which is tied to a different architecture) can help you reach what you're trying to accomplish?
I use comfy and SDXL on a laptop so you don't have to worry about the "requirements".
Also, the bigger the model (sd1.5 < SDXL < Flux), the more your prompt needs to be "natural language" and less "keywords oriented".
This is something I've learnt the hard way, hating on XL because I was not really... Speaking it's language.
@oddKoodoo about that: Just try to download both the full-sized version and a pruned one of the same model and generate two images with the same settings. ✨😌 You'll get the same exact image twice.
@Arthemy I've tried XL-version earlier and I don't remember why, but I didn't liked it, mb it has less pronounced style... Actually I've got interested in v1.5 model in first place because upscale is quick and brings more unexpected details. While XL-upscale is really VRAM-expensive and slow... However, I've downloaded XL-version and I'll try it again.
@Arthemy I've compared pruned and full models earlier and thought that it just I can't see the difference, but in cases of some prompts and settings changes can be seen...
Hi! Thank you for an awesome model!
Βut could you advise, please, why I'm always getting grey and desaturated results while inpaintings? Any tips to fix it? Many thanks!
I suggest you to download the "BakedVAE" version, that should solve your problem! :D
This is by far one of the most stable models out there. Thank you!!!
Thank you very much! I can't wait to see what you're going to create with it! ✨
The stability and baked in beautiful animated style is ideal for inpainting pony images. This model gives me more creative options. Thank you for your work.
You welcome! I remember to have worked a lot on this model back in the days and I'm happy that it's still able to produce quality content to inspire and help people to realize their ideas. I'm glad to hear positive feedback like this :3
I love this model, but I cant get it to do the hands right. (Even more difficult than on other model) Any tips?
Great model!