CivArchive
    Otherworld IDXL/SDXL Hybrid - v4.0
    NSFW
    Preview 76538993
    Preview 76667236
    Preview 78089124
    Preview 76503615
    Preview 76544845
    Preview 76502798
    Preview 76505177
    Preview 76551155
    Preview 76554081
    Preview 76503993
    Preview 76508688
    Preview 76503048
    Preview 76502889
    Preview 76503557
    Preview 76508590
    Preview 76508982
    Preview 76507742
    Preview 76656803
    Preview 79657701

    If you appreciate my work and want to see more updates and models from me in the future, please like, follow, share the art you create with my models, and tip me some Buzz, your support is what motivates me to continue this work, thank you.

    This model has been baking for months, it is a merge made from other merges I made while working on my other models, it is a hybrid of Illustrious and SDXL focused on balancing realism with Illustrious' built in knowledge of anime game and cartoon characters

    What's in the mix? I often keep models I have made that were 'merge fodder' for my other models, and it is mostly comprised of those, with the additions of EpicRealism LastFrame SDXL and CognitiveRiver v1 IDXL

    CFG 2.5-6
    Euler A, DPM++ 2M Karras, DPM++ 3M SDE Karras
    Steps 15-60


    Positive: RAW:1.4, ultrarealistic:1.4, masterpiece:1.4, (Your prompt here) , best quality, amazing quality, very aesthetic:1.6, absurdres, 8K, newest, focused subject, depth of field, shallow depth of field, dramatic cinematic lighting, moody shadows, low-key lighting, DSLR, realistic, very awa, cosplay photo, year 2024, highres:1, skin texture, facial features, photography shot with Canon EOS 5D Mark IV, skin details, cinematic film still, sharp focus, intricate details, highly detailed, hyperrealistic art, vignette, high budget, bokeh, cinemascope, moody, epic, gorgeous, film grain, grainy

    Negative: bad quality, worst quality, low quality, poor quality, worst detail, lowres, blurry, out of focus, pixelated, jpeg artifacts, scan artifacts, lossy, ugly, boring, plain, censored, watermark, logo, signature, artist name, sketch, cartoon, drawing, anime, illustration, fake, phony, artificial, CGI, 3D, 3D render, digital model, simplistic, disfigured, deformed, mutation, mutated hands, extra limbs, extra fingers, extra hand, extra arm, extra leg, merged fingers, fused fingers, fused limbs, merged body parts, deformed hand, deformed finger, imperfect hand, mutilated, deformity, bad anatomy, poorly drawn face, poorly drawn eyes, mutated face, ugly face, disfigured face, deformed face, deformed eyes, out-of-frame, animated

    As always, have fun, and please create responsibly.

    Description

    Added in Lucent XL, SmoothmixSemireal, and did a DARE merge with Otherworld 1.0
    Big improvement on Skin texture
    Slightly more realism
    Improved conceptual knowledge

    FAQ

    Comments (8)

    JellaiJun 4, 2025· 1 reaction
    CivitAI

    I think your model looks great, but I'm not getting aanywhere near the quality you are. I've used your settings/prompts/loras/seeds exactly and all of my results look like digitally painted plastic, while yours has subsurface scattering and way more life to it.

    Now I've gone over your prompts carefully to see if there's a hidden lora or something weird, but the only thing that popped out to me was that at the end of your prompt, there was a section starting with "[Enhance]" and ending with "[/Enhance]". I'm guessing this is where things were fixed up? What workflow did you use for this? what is it? I'd love to match your outputs. I think your model is unique. Thanks!

    Urameshi
    Author
    Jun 4, 2025

    I use A1111 webui, so double check your comfy settings make sure you have clip skip 2, make sure it isnt doing anything extra (including using a VAE) cause I bake in my VAEs in all my models, if all else fails and you still cant get the same output in comfy Id recommend using A1111 webui for my models. As for the enhance thing thats an enhancement prompt an AI chat model gave me, thats how the AI wrote it out and Ive gotten good results with it as it is so I never bothered to change it and I dont really know if the [enhance] [/enhance] part does anything it may just be how the model was presenting me the prompt

    JellaiJun 4, 2025

    @Urameshi Ah interesting. Thank you! Yeah, in ComfyUI, I just tried your Squirtle image using "DPM++ 3M SDE" and all the other settings identical, and it was like it had the wrong/no vae, showing as a garbled mess. I used Euler and that fixed it, but again, the image quality looked plain and plastic. I am indeed using your model's vae.

    I used to use A1111 all the time, and loved it, but it hasn't been developed in years, and I just don't know what security issues are in there now. I moved on. That's cool that the model looks good there though. I just think the lack of Comfy support is likely why you aren't getting many images in the gallery from people. You've clearly made a really good A1111 model.

    Thanks for the response.

    Urameshi
    Author
    Jun 4, 2025· 1 reaction

    @Jellai I'd imagine there must be a way to use my models in comfy cause my illustrious realism has a ton of user generated images and videos and Im pretty sure Ive seen some workflows in a few of those, if you could find a workflow that works with that model I would think it would also work with this model, I wish I could help ya more but A1111 has the merging extensions I like to use so Ive just stuck with it, I did try comfy but just didnt care for the feel of it, and didnt wanna lose my merger extensions, thank you for the feedback, on your security concern, maybe forgeUI would work? its a fork of A1111 and Im pretty sure thats still updated, if its not you could see if there is a fork thats still being regularly updated

    JellaiJun 5, 2025· 1 reaction

    @Urameshi Yeah, after your previous comment, I started setting up Forge, specifically for your model. I do miss the old A1111 interface sometimes, so I feel kinda nice about coming back home after years away from it. Oh and to be clear, your model does load and generate images in Comfy, but the results are just strangely lower quality using the same prompts and setups, though it's incompatible with that one sampler for some reason. It's not lower quality in a way that looks broken. Just much less life. Anyway, we'll see if Forge works out. Thanks again.

    Urameshi
    Author
    Jun 5, 2025

    @Jellai Well I hope at somepoint I can find out why comfy is making lower quality outputs with my model, I hope its just your version of comfy maybe something you have going on with your comfy like an addon, or workflow, or something stupid like the schedule type your using, or that comfy is selecting for you, I use automatic with euler A, maybe you were using karras or beta or something else? or maybe comfy was selecting a different schedule type with automatic than what A1111 selects, Im glad you brought this to my attention cause I would have never known, I assumed it would be working the same in comfy, but now Im definitely curious as to what the issue could be, at least in the meantime using forge should give the same quality Im getting with a1111, but Imma have to find some people that are very knowledgeable with comfy and see if they can figure out what might be going on

    Noob_eeJun 6, 2025

    Have you tried upping the "steps" in comfy? sometimes that helps or lowering the CFG. Sometimes that helps. Best regards.

    JellaiJun 10, 2025

    @Noob_ee I have. I was able to match v40's look, but no luck with v30 yet. The result is just so different. Thanks for your suggestions.

    Checkpoint
    Illustrious

    Details

    Downloads
    877
    Platform
    CivitAI
    Platform Status
    Deleted
    Created
    5/15/2025
    Updated
    4/21/2026
    Deleted
    4/16/2026