This is my flagship Incursio's Meme Diffusion (SDXL, Pony). I find it's in line with what the community on Civitai likes producing. It's pretty versatile while also having an artstyle that's unique in image generation. NSFW compatible. This is the XL version of my Incursio's Meme Diffusion.
Refer to the version details for Changelog.
V2.7:
This Checkpoint is trained on the awesome Basemodel AutismMix_confetti.
V2.1:
This Checkpoint is trained on the awesome Basemodel Pony Diffusion V6.
This is in Collaboration with @Anzhc . Check out his AAAAustism 馃槈
V1.6:
This Checkpoint is trained on the awesome Basemodel AutismMix_confetti.
Thanks to @Autismix_anon for allowing this merge. Much love buddy 鉂わ笍
This is in Collaboration with @mfcg. Check out his MCFG PDXL 馃槈
The only prompt you ever need is this positive prompt:
score_9, score_8_up, score_7_upV1.6 can use source_anime and V2.1 can use detailed as an additional tag. veins is a pretty strong tag, as well as toned in V2.1.
No need for any negatives. Please dont use any or at least only use negatives, that you dont want in your current image. I sometimes use monochromenegative only. I'm sick of seeing schizo negatives and schizo positives. I also recommend to not use any () or with weights like (:1.2) on any tags. It can produce some artifacts on your images. To know more about this, check this Article. Anything else for tag related stuff, see Basemodel Pony Diffusion V6.
I also recommend to use the Extension Adetailer, because eyes may not be sharp due to burns.
Settings for generating Images for V1.6:
Sampler: Euler a
Steps: 24
CFG 6.0
Base Resolution 1024x1024
Settings for generating Images for V2.1:
Sampler: Restart
Steps: 14-18
CFG 4.0-6.0
Base Resolution 1024x1024
Settings for generating Images for V2.7:
V2.7 Can use the same settings from v1.6 and v2.1. Style and Details change depending on the Sampler.
Recipe V1.6:
Merge to Checkpoint: TC V Style LoRA PD on Weight 0.75 + AutismMix_confetti.
Recipe V2.1:
Merge to Checkpoint: TC V Style LoRA PD V2 on Weight 1.0 + Basemodel Pony Diffusion V6.
Recipe V2.7:
Merge to Checkpoint: TC V Style LoRA PD V3 on Weight 0.65 + AutismMix_confetti = Meme Diffusion v2.3 (not released).
Additional Merge with block weights and adjustment:
incursiosMemeDiffusion_v23 x (1-alpha) + incursiosMemeDiffusion_v21 x alpha (0.2,0.017,0.434,0.052,0.485,0.038,0.652,0.206,0.421,0.032,0.144,0.073,0.27,0.249,0.331,0.849,0.653,0.847,0.979,0.065,0)
Adjustment: -0.38,0.55,0,0.59,-2.74,0,0,0.92
Everything else is specified in the image tags. Make sure to leave a review or a comment if you got good use out of this Mix.
How to use Checkpoints's in auto1111:
Download Checkpoint
Copy the file to
stable-diffusion-webui/models/Stable-diffusionSelect your Checkpoint
Have Fun!
Description
Style is now stronger towards v1.6
Same training as v2.1
Adjusted Merge and block weights
Depending on Sampler (Euler a, Restart) results may differ.
Better night/dark enviroment images
Leaning towards 2.5D without having cursed smile
FAQ
Comments (60)
V2.7 should now be better overall. It now leans more towards v1.6 Style while maintaining a bit less details from v2.1.
Color problem should be now fixed with this one.
On-Site generator will be switching to the new version soon and old versions will be closed for the generator, unless there is demand. Any feedback is appreciated.
Talk about timing, i was just messing around with 2.1, im curious to see how much the new version holds up.
Nice Job Man :)
where doro? Day ruined
This Model is not working anymore.
I have do one Picture of Princess Luna at 4:54am who has worked fine (I have remix it from this Picture on Version 2.1).
Now I wanted to do another with her Sister Celestia after 5 am and I just get the Error:
Failed to generate
Resource urn:air:sdxl:checkpoint:civitai:314404@574756 is not enabled for generation.
The Generator for this Model changed to v2.7 for the time being, because civitai doesn't have enough gpus to support all versions. If the demand for each version is higher than v2.7, then I will swap back
@FallenIncursio聽I have try the new version but always if I use the remix of pictures before the tool switch to pony diffusion or a broken version of this model, so I needed to readd it every time, I have fix it by just changing the prompt and I have create some realy nice sfw pictures of the mlp main figures (Luna, celestia, cadance, flurry, chrysalis, the mane twilight, fluttershy, pinkie, rainbow, aj, rarity, starlight & sunset and the specials nightmare moon and twilight velvet.
I guess I will post them today.
Actual I just thinking to ad more char in the mix.^^
Woah, this model improved by a lot! Loving it to use on Civitai generator!
yeah no, this still doesnt feel right. Please leave 1.6 as a legacy option. Its not negatively impacting anything by being there.
favor, dejen la versi贸n 1.6 como una opci贸n heredada
Why did they remove v 1.6 I don't like it 2.7 it looks strange I believe the images
I miss the old checkpoints
Seems like people are rly attached to V1.6. It's now open for the Generator again, but I rly courage people to use V2.7 instead, because of the overall improvements in the ckpt (like better night/dark images, better details, etc...).
Also, pls dont blame the Style of the Ckpt itself, when it gets used with 50 Style LoRA's and 10 Embeddings. It's best to use 1-4 negative prompts and controlled positive Prompts with a handful of LoRA's.
As a DIEHARD lover of the original meme diffusion. I think most people鈥檚 issues with it isn鈥檛 personal bias. The new ones you feel SUBSTANTIALLY darker than the original one which doesn鈥檛 always work.
These are two images, same description, same seed. The only difference is 1.6 and 2.7. https://civitai.com/posts/5424034
The difference is night and day. But I鈥檓 just one person. Maybe it is personal bias for everyone else.
As bob908 said, it's a matter of DAAAARKNEEEESS.... Just kidding XD But memes aside, not only the 2.1/2.7 versions feel a lot darker (2.7 is more balanced, but still....), I would also add that the contrast is a lot sharper in those as well (again 2.7 being more subtle, tho noticeable), which loses part of the charm of the OG 1.6 version.
I did a test with two pics being done in the exact same way with both 1.6 & 2.7, and while the model rendered them differently for being, well, different versions and thus "different" models (for example one gave the chars toeless socks, the other didn't). One thing I noticed is that the characters on the new version had a much harder borderline than the OG one. To put in perspective. The OG 1.6 feels like a 2.5D cartoonish style while 2.7 feels like using Superhero Comic style. Both lean into the semirealistic vibe with a more 2D bias but the light and contrast balancing in both feel way different.
To be honest I don't think it's a bad idea to keep both available. I'll say each feel better for different things cause as you mentioned. I did noticed that 2.7 handles details dramatically better than 1.6. For example in those pics the chars had spiked collars and wristbands/anklets and 1.6's designed them looking very basic but on 2.7 they felt like the ones you'll find in a heavy metal accessories shop IRL lol
I'm not gonna ignore the fact that I personally prefer this model in the 1.6 version cause is the one I remember when I think of it, but I wouldn't say I dislike the other two versions either when looking for a more dramatic, stylized design.
Thats some feedback I can work with. Much love. Next version hopefully much better. Any additional feedback is welcome. 鉂わ笍
@FallenIncursio聽Anytime ^^
I'm currently making a series of pics directly comparing the 3 versions and maybe making a hi-res of the one I liked the most, cause why not hehe. I'll try to mention the things I like the most about each version. That is if SD starts to behave. Cause sometimes it does everything but what I tell it to do -w-
I just made the post comparing the 3 of them. I tagged you in it @FallenIncursio so you could take a look. I used the post tagging feature thingy so, not sure how that works exactly but I wanted to let you know just in case XD
There's a description on what I noticed in them that I also saw in the other renders I did with them. As well as a render with the same 1.6 version but on the online generator to compare how different Stable Diffusion and CivitAI's generator looked like. Gotta say, it's scarily different .w.
NGL, i've tried many checkpoints but none meassure up to this one. For some reason i can't find something better than this for SDXL and PONY images.
one of the best, hard to find a substitute.
I keep trying anime checkpoints and coming back to this one. I always get high quality and decent anatomy on this one
As amazing a checkpoint as it is, sometimes I struggle with getting the hair color of characters right, does this model prefer black hair the most?
No, it doesnt struggle with hair colors. Something else is breaking your image
I see, thank you for responding. I am gonna redouble the efforts on prompting then, I love this model lol.
Civitai having a shortage of GPU's currently, so im limited by the amount of Checkpoints I can put online for the generator.
V2.7 currently has the highest usage, so I had to turn off V1.6
I would recommend to check out PaSanctuary.
Will you eventually re-enable V1.6? It was one of the best checkpoints i've tried, one of the best imo
@unempty聽I would reopen it, but its currently looking dry in Civitai. I have no idea, when its going to be open again
PaSanctuary is not a pony system, so I gave up because I could not use lora.
Seems to be happening to other creators as well. All their models straight up are becoming unavailable to be used on the generator. Gosh I swear this website's system gets worse each day -w-
2.7 is very dark, better 1.6
Hope you can keep it on Huggingface so others can use offline if they can't on CivitAi
Love the new version however the images are too dark, how to remedy this?
Such a fantastic and pivotal model.
Each version captures how popular styles changed with PDXL over time.
In my opinion V1.6 is the most iconic version.
While V2.1 has the best detail, textures, and color.
Finally V2.7 feels somewhere between V1.6 and 2.1 in its look and feel.
when ilustrious version?
This super model is not available in website anymore?
This is all because of the stupid auction system, which causes those checkpoints that did not make it into the top 200 at the end of the auction to drop out of online generation for a week until this checkpoint gets into the top 200. The author of this checkpoint is not to blame.
@normad11聽Thank you :) I blamed no one, just needed to understand (I bid for this but i had not understand right the system :/ )
@Zail聽https://civitai.com/articles/12904 all information 馃槕
@normad11聽Yes, read it, not understood it xD But now it's ok cause of you
Check again! Help me keep it alive on the next auction!
Incursio not available for generation :(
It is now! Bid with me the next time it is up on auction! :D
Way was this disabled
Both Incursio's Meme Diffusion and Prefect Pony, two amazing popular checkpoints with a combined total of over 7 million generations, have been disabled due to the new, incomprehensibly stupid auction system. In other words, it's CivitAI slowly destroying itself with one ridiculous decision after another.
@MakiZeninHasNoScars聽Me: "Should I give another chance to CivitAI after all this time to see if they fixed their mess?"
CivitAI: "I present you the Auction System, in which about half your favorite checkpoints that were available from back when we had a limited use of GPUs are now voted on the chance to see if they make it to the top so they're available"
Me: "Well, I guess I'll stick with Stable Diffusion locally as I have been until now then...."
@MakiZeninHasNoScars聽well the a stupid choice
I've tried countless models; none do skin textures and nuanced, anatomically correct underlying musculature like this one.
Of course, left to its own devices, it really, really, REALLY wants to draw mountains of meat that would make Hulk Hogan and Lizzo look small, so half the prompt is reigning it in, but small price to pay!
Finaly my favorite model is back. My pics without your model was very garbage and the style was horrify
Yay Incursio is back!!!
Plese, we need an Illustrious version of this model!
Wonderful. Can't use the only model I care about for another week. This site can die in a fire for all I care.
So is this just not usuable on the website?
Really hope this model comes back
Illustrious version, please!
Got some of the best generations I ever got yesterday but suddenly they looked awful today. Turns out this got disabled for website generation :(
Please be back soon!
One of the best Pony checkpoints I've ever used, that's for sure.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.















