AniVerse XL

My personal page for donation (and where you can buy my model in early preview): <a target="_blank" rel="ugc" href="https://ko-fi.com/samael1976">Samael1976 Ko-fi page</a> What New: This is an update that is totally different from the previous versions. So if you like the Swiss Army knife that is version 2.0 (in the sense that it can do more or less all the styles), stick with that version.I've been feeling the call to update AniVerse XL for a long time, it had become a nagging problem in my head.I tried to make the output of the images uniform with the classic style of AniVerse, even without having to use embeddings.I think I partially succeeded, it certainly needs other future refinements, but this is a first step. &lt;-----------------------------------------------------------------------------------&gt;I also tried with embeddings and I can tell you that: <a target="_blank" rel="ugc" href="https://civitai.com/api/download/models/480651?type=Model&amp;format=SafeTensor">unaesthetic_bp5</a><ul><li>creates a bit of problems with the hands</li><li>heavily retouches the "creativity" of the image.</li><li>Also in long focal lengths (full body shot) the faces do not render as well as they should</li></ul><a target="_blank" rel="ugc" href="https://civitai.com/api/download/models/972770?type=Negative&amp;format=Other">CyberRealistic_Negative_PONY-neg</a><ul><li>it is more adherent to the original image and works very well.</li><li>Although it tends to gloss over images a bit too much and make them look flat, and removes a little bit of details.</li><li>it tends to make faces a little too rounded (for my taste)</li><li>It works much better with Euler Max than DPM++2M</li><li>Perfect for making male figures</li></ul> -&gt; Therefore, my advice is to continue reading these few lines and use the negative prompt that I suggest below &lt;- &lt;-----------------------------------------------------------------------------------&gt;About samplers with Karras like scheduler:<ul><li>DPM++2M: this is my favorite, both for colors and details, it makes the image more towards 2.5D, great for portraits and cowboy shots.</li><li>Euler Max: It has less details, but it renders better at the face level when using the "full body shot" prompt and tends to have a greater proximity to 2D.</li></ul>You decide what to choose based on your needs.&lt;-----------------------------------------------------------------------------------&gt;<ul><li>These are the recommended prompts and parameters: Clip skip: 2 Width: 768 Height: 1344 CFG: 5.5 Steps: 30 Sampling: DPM++ 2M o Euler Max Scheduler: Karras Trigger Word: 4n1v3rs3   EXAMPLE OF GENERAL PROMPT:   POSITIVE: (Type of Shoot), (Subject), adult, the description, (background), more details, hyperdetailed:1.15, 4n1v3rs3   EXAMPLE: Portrait, Zelda, adult, cute, seductive, innocent, light smile:0.3, plump lips, slender body, ankle-length pink dress, jeweled tiara, golden shoulder pads, a gold chain as a necklace, vibrant, fantasy, epic, heroic, cave background, depth of field, dynamic angle, fashion photography, sharp, hyperdetailed:1.15, 4n1v3rs3   NEGATIVE: worst quality:1.4, low quality:1.4, front light, grayscale, ugly, fat, wide hips, curvy, child, young, kid, bad hands, interlocked fingers, extra finger, missing finger, fused fingers, bad anatomy </li></ul>&lt;-----------------------------------------------------------------------------------&gt;PS: now, although always with some limitations, you can do nude poses PPS: Sorry, I know the hands aren't the best, I'll try to fix them with the next version

aniverseXL_v30.safetensors

v3.0

v4.0

v2.0

v1.0

⬇Read the info below to get the high quality images (click on show more)⬇&gt;&gt;&gt; UPLOADING/SHARING MY MODELS OUTSIDE CIVITAI IS STRICLY PROHIBITED* &lt;&lt;&lt;The only authorized generative service website are:• <a rel="ugc noopener noreferrer nofollow" href="http://Mage.Space" target="_blank">Mage.Space</a> = That's help me a lot to buy a new PC• Also in collaboration with <a rel="ugc noopener noreferrer nofollow" href="http://Shakker.ai" target="_blank">Shakker.ai</a>This model is free for personal use and free for personal merging(*).For commercial use, please be sure to <a rel="ugc noopener noreferrer nofollow" href="https://ko-fi.com/samael1976" target="_blank">contact me </a>(Ko-fi) or by email: samuele[dot]bonzio[at]gmail[dot]comLeaderboard of my 3 best supporter:<a rel="ugc noopener noreferrer nofollow" href="https://ko-fi.com/samael1976/leaderboard" target="_blank">https://ko-fi.com/samael1976/leaderboard</a>Aniverse XL - make the impossible possible!This is a long shot project, I’d like to implement something new at every update!The name is a merge of the two words: Animation and Universe (and a word pun: Any+Universe -&gt; Anyverse -&gt; Aniverse)-&gt; If you are satisfied using my model, press on ❤️ to follow the progress and consider leaving me ⭐⭐⭐⭐⭐ on model review, it's really important to me!Thank you in advance 🙇And remember to publish your creations using this model! I’d really love to see what your imagination can do!Recommended Settings:<ul><li>Excessive negative prompt can makes your creations worse, so follow my suggestions below!</li><li>Before applying a LoRA to produce your favorite character, try it without first. You might be surprised what this model can do!</li></ul>A1111 my settings:I run my Home PC A1111 with this setting:<ul><li>set COMMANDLINE_ARGS= --xformers --skip-torch-cuda-test --no-half-vae (if you have low VRam, try to add --medvram-sdxl or --lowvram that can help you, but it slow down the image creations)</li></ul>if you can't install xFormers (read below) use my Google Colab Setting:<ul><li>set COMMANDLINE_ARGS= --disable-model-loading-ram-optimization --opt-sdp-no-mem-attention --no-half-vae (if you have low VRam, try to add --medvram-sdxl or --lowvram that can help you, but it slow down the image creations)</li></ul>My A1111 Version: version: v1.9.3  •  python: 3.10.11  •  torch: 2.1.2+cu121  •  xformers: <a rel="ugc noopener noreferrer nofollow" href="http://0.0.23.post" target="_blank">0.0.23.post</a>1  •  gradio: 3.41.2  • If you want activate xformers optimization like my Home PC (<a rel="ugc noopener noreferrer nofollow" href="https://aipromptshome.com/fix-no-module-xformers-stable-diffusion-automatic-1111/" target="_blank">How to install xFormers</a>):<ul><li>In A1111 click in "Setting Tab"</li><li>In the left coloumn, click in "Optimization"</li><li>in: "Cross attention optimization" select: "xformers"</li><li>Press in "Apply Settings"</li><li>Reboot your Stable Diffusion</li></ul>If you can't install xFormers use SDP-ATTENTION, like my Google Colab:<ul><li>In A1111 click in "Setting Tab"</li><li>In the left coloumn, click in "Optimization"</li><li>in: "Cross attention optimization" select: "sdp-no-mem - scaled dot product without memory efficient attention"</li><li>Press in "Apply Settings"</li><li>Reboot your Stable Diffusion</li></ul>How to emulate the nvidia GPU follow this steps:<ul><li>In A1111 click in "Setting Tab"</li><li>In the left coloumn, click in "Show all pages"</li><li>Search "Random number generator source"</li><li>Select the voice: "NV"</li><li>Press in "Apply Settings"</li><li>Reboot your Stable Diffusion</li></ul>If you use my models, install the <a rel="ugc noopener noreferrer nofollow" href="https://github.com/Bing-su/adetailer" target="_blank">ADetailer</a> extension for your A1111.Navigate to the "Extensions" tab within Stable Diffusion.<ul><li>Go to the "Install from URL" subsection.</li><li>Under "URL for extension's git repository" put this link: : <a rel="ugc noopener noreferrer nofollow" href="https://github.com/Bing-su/adetailer" target="_blank">https://github.com/Bing-su/adetailer</a></li><li>Click on the "Install" button to install the extension</li><li>Reboot your Stable Diffusion</li></ul>How to install Euler Max Sampler:<ul><li>In A1111 click in "Extensions Tab"</li><li>click in "Install from URL"</li><li>Under "URL for extension's git repository" put this link: <a rel="ugc noopener noreferrer nofollow" href="https://github.com/licyk/advanced_euler_sampler_extension" target="_blank">https://github.com/licyk/advanced_euler_sampler_extension</a></li><li>Once installed click in: "Installed" Tab</li><li>Click in "Apply and quit"</li><li>Reboot your Stable Diffusion</li><li>Now at the end of the list of the sampler, you have the new sampler.</li></ul>

My personal page for donation (and where you can buy my model in early preview): Samael1976 Ko-fi page

What New: This is an update that is totally different from the previous versions.
So if you like the Swiss Army knife that is version 2.0 (in the sense that it can do more or less all the styles), stick with that version.

I've been feeling the call to update AniVerse XL for a long time, it had become a nagging problem in my head.

I tried to make the output of the images uniform with the classic style of AniVerse, even without having to use embeddings.

I think I partially succeeded, it certainly needs other future refinements, but this is a first step.
<----------------------------------------------------------------------------------->

I also tried with embeddings and I can tell you that:


unaesthetic_bp5

creates a bit of problems with the hands

heavily retouches the "creativity" of the image.

Also in long focal lengths (full body shot) the faces do not render as well as they should

CyberRealistic_Negative_PONY-neg

it is more adherent to the original image and works very well.

Although it tends to gloss over images a bit too much and make them look flat, and removes a little bit of details.

it tends to make faces a little too rounded (for my taste)

It works much better with Euler Max than DPM++2M

Perfect for making male figures


-> Therefore, my advice is to continue reading these few lines and use the negative prompt that I suggest below <-
<----------------------------------------------------------------------------------->

About samplers with Karras like scheduler:

DPM++2M: this is my favorite, both for colors and details, it makes the image more towards 2.5D, great for portraits and cowboy shots.

Euler Max: It has less details, but it renders better at the face level when using the "full body shot" prompt and tends to have a greater proximity to 2D.

You decide what to choose based on your needs.

<----------------------------------------------------------------------------------->

These are the recommended prompts and parameters:
Clip skip: 2
Width: 768
Height: 1344
CFG: 5.5
Steps: 30
Sampling: DPM++ 2M o Euler Max
Scheduler: Karras
Trigger Word: 4n1v3rs3
 
EXAMPLE OF GENERAL PROMPT:
 
POSITIVE: (Type of Shoot), (Subject), adult, the description, (background), more details, hyperdetailed:1.15, 4n1v3rs3
 
EXAMPLE: Portrait, Zelda, adult, cute, seductive, innocent, light smile:0.3, plump lips, slender body, ankle-length pink dress, jeweled tiara, golden shoulder pads, a gold chain as a necklace, vibrant, fantasy, epic, heroic, cave background, depth of field, dynamic angle, fashion photography, sharp, hyperdetailed:1.15, 4n1v3rs3
 
NEGATIVE: worst quality:1.4, low quality:1.4, front light, grayscale, ugly, fat, wide hips, curvy, child, young, kid, bad hands, interlocked fingers, extra finger, missing finger, fused fingers, bad anatomy

<----------------------------------------------------------------------------------->

PS: now, although always with some limitations, you can do nude poses
PPS: Sorry, I know the hands aren't the best, I'll try to fix them with the next version

Files with SHA256: 5d4a2061fb93dc68fde97761e0b6d258d133a5a793712a685361079634378b71