santa hat
deerdeer nosedeer glow
Sign In
home
models
images
videos
posts
articles
bounties
tools
challenges
events
shop

Copax TimeLess

7.3k
147.1k
2k
Early Access
Updated: Dec 17, 2024
base modelrealismrealisticstylescopaxrealisticxlsdxl1copax timeless xl
Verified:
GGUF
Type
Checkpoint Trained
Stats
14
Reviews
(17)
Published
Dec 17, 2024
Base Model
Flux.1 D
Hash
AutoV2
8FE673AB7C
Oni Badge
Copax's Avatar
Copax
License:
FLUX.1 [dev] Non-Commercial License
The FLUX.1 [dev] Model is licensed by Black Forest Labs. Inc. under the FLUX.1 [dev] Non-Commercial License. Copyright Black Forest Labs. Inc.
IN NO EVENT SHALL BLACK FOREST LABS, INC. BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH USE OF THIS MODEL.

_________________________

Check my models on Mage:

_____________________________

--- XPlus 3/ XPlus MIX 3/ XPlus 3 Dark ---

  • The most noticeable difference between the XPlus 3 and XPlus MIX 3 models is the lighting effect. The XPlus 3 delivers an absolute black tone more effectively, while the XPlus MIX 3 excels at warm tones, providing softer light that highlights facial and skin details more clearly.

  • For example, in a photo taken in a dark room, the XPlus 3 produces a deeper black background, whereas the MIX 3 makes the face appear brighter and more evenly toned. In outdoor photos under sunlight, the MIX 3 creates a more realistic and natural effect.

  • Additionally, there are some differences in hair, skin, and material rendering. Finally, the XPlus 3 Dark model is undoubtedly the darkest version, evoking lighting and atmosphere similar to images set in dungeons. The Dark 3 also introduces more noise, making it ideal for dark art, horror themes, and images that require a gloomy atmosphere.

_____________________________________________

Copax Timeless ver XPlus 3 (Canny)

This is a model combination of the XPlus 3 model and FLUX.1-Canny-dev-lora.

Introducing FLUX.1 Tools by BlackForestLabs: https://blackforestlabs.ai/flux-1-tools/

  • Note: ComfyUI is required to use this model.

Recommend

ComfyUI workflow for the model: https://civitai.com/articles/9065

  • Steps: 20

  • Distilled CFG Scale: 4.5

  • CFG : 1

  • Diffusion in Low Bits: float8-e4m3fn

  • Sampling: Euler simple

  • Nprompt: none

  • Vae: flux vae (must have)

  • Clip: clip_l.safetensors and t5xxl_fp8_e4m3fn.safetensors (must have)

_________________________________

Copax Timeless ver XPlus 2 (base train)

Note: If you only want to create images, it's not necessary to download this version; you can use other versions of the model.

  • I want to clarify about the XPlus 2 (base train) model, which is the base model used for training the XPlus 3 version.

  • Initially, I planned to release the XPlus 3 23GB version as the base model used for training.

    However, test results showed that the XPlus 3 23GB version is not as stable as the XPlus 2 (base train).

  • Therefore, I will not release the XPlus 3 23GB version and will instead release the XPlus 2 (base train) version.

_________________________

Copax Timeless ver XPlus 1/2/3

Hello everyone,

  • This model extends the capability to generate NSFW content for flux1-dev-fp8.

  • This version focuses on Fix some errors in XPlus_1 such as nipples appearing excessively..etc.... Sex poses, penis do not work well in the current version.

  • Additional more character data. Added male character data.

  • Currently male nude with penis does not work well in this update

If you enjoy the work I’m doing and want to see more updates in the future, please consider supporting me by leaving a like, review, tip, or making a donation. Your support will be a great motivation for me to continue bringing new models in the coming times.

Thank you so much for always being there and supporting me!

Recommend

  • Steps: 20

  • Distilled CFG Scale: 3.5

  • CFG : 1

  • Diffusion in Low Bits: float8-e4m3fn

  • Sampling: Euler simple

  • Nprompt: none

  • Vae: flux vae (must have)

  • Clip: clip_l.safetensors and t5xxl_fp8_e4m3fn.safetensors (must have)

_________________________________

Copax Timeless ver FluxReal (10 step)

This is an Flux model focused on a photography style with fast image generation (only 10 steps needed).

Recommend

  • Using forgeUI to run the model

  • Steps: 8~10

  • Distilled CFG Scale: 1

  • CFG : 1

  • Sampling: Euler simple

  • Diffusion in Low Bits: Automatic

  • Vae: flux vae (must have)

  • Clip: clip_l.safetensors and t5xxl_fp8_e4m3fn.safetensors (must have)

____________________________________

Copax Timeless ver Photorealism (SDXL 8 step)

This is an SDXL model focused on a photography style with fast image generation (only 8 steps needed). If you prefer creating images quickly within a few seconds or want a model that matches your hardware specifications, this could be a suitable choice.

Recommend

Steps: 8,

Sampler: DPM++ SDE,

Schedule type: Karras,

CFG scale: 2,

You can also using After Detailer for face

____________________________________________

Copax Timeless ver XPlus_MIX

XPlus_MIX models focus on photography, while StyleMix concentrates on style art, 3D, and painting

I plan to continue developing and enhancing this model in the future. Training a Flux model is very costly, so I've launched an Early Access option in the hopes of gaining support to fund the development of future versions.

If you enjoy the work I’m doing and want to see more updates in the future, please consider supporting me by leaving a like, review, tip, or making a donation. Your support will be a great motivation for me to continue bringing new models in the coming times.

Thank you so much for always being there and supporting me!

Recommend

  • Using forgeUI to run the model

  • Steps: ~20

  • Distilled CFG Scale: 3.5

  • CFG : 1

  • Sampling: Euler simple

  • Diffusion in Low Bits: Automatic

  • Vae: flux vae (must have)

  • Clip: clip_l.safetensors and t5xxl_fp8_e4m3fn.safetensors (must have)

________________

StyleMix_2

Although it’s still a bit early, some of you might already be starting to plan designs, print cards, clothing, and banners for Christmas and New Year themes.

That's why I've prepared this model, which focuses on a cozy style, making it suitable for Christmas and New Year themes.

Recommend

  • Using forgeUI to run the model

  • Steps: ~20

  • Distilled CFG Scale: 3.5

  • CFG : 1

  • Sampling: Euler simple

  • Diffusion in Low Bits: Automatic

  • Vae: flux vae (must have)

  • Clip: clip_l.safetensors and t5xxl_fp8_e4m3fn.safetensors (must have)

____________

Copax Timeless ver XPlus_2(GGUF Q8)/(GGUF Q6)/(GGUF Q4)


This is the GGUF version of model X Plus_2. This model has a file with the extension .gguf Use comfyUI to run this model.

The model's workflow you can download here: ComfyUI-GGUF simple workflow

!!Note: Version GGUF Q4 uses less vram but has a higher error rate than version GGUF Q8. You may have to regenerate a few times to get the desired results. If you want better quality, you can use GGUF Q8 or other versions.

Version GGUF Q6:

Although I haven't had much experience yet, based on the initial test results, I find the GGUF Q6 version to be a very interesting model. 

It delivers surprisingly high-quality material details, skin, and character details. 

I used the wildcard Processor when generating images on ComfyUI, so it seems that the prompt doesn't display when viewing the image.

Thank you pkmngotrnr, for helping me convert the model to the gguf version and providing me with usage instructions.

______________________________________________________

StyleMix_1(nf4 LowVram)

This version has a capacity of 6gb and does not include Clip and VAE, it uses less Vram.

If you want better quality, you can use version StyleMix_1(fp8) or StyleMix_1(fp16)

Recommend

  • Using forgeUI to run the model (must have) ( Currently comfyUI node NF4loader cannot run this model)

  • Steps: ~20

  • Distilled CFG Scale: 3.5

  • CFG : 1

  • Sampling: Euler simple

  • Diffusion in Low Bits: bnb-nf4

  • Vae: flux vae (must have)

  • Clip: clip_l.safetensors and t5xxl_fp8_e4m3fn.safetensors (must have)

__________________________________

StyleMix_1

The StyleMix models will focus on art styles, with StyleMix_1 generating images with softer strokes leaning towards anime and painting.

The StyleMix_1 model is SFW and is a Flux Dev FP16 version with a size of 23GB.

I plan to continue developing and enhancing this model in the future. Training a Flux model is very costly, so I've launched an Early Access option in the hopes of gaining support to fund the development of future versions.

If you enjoy the work I’m doing and want to see more updates in the future, please consider supporting me by leaving a like, review, tip, or making a donation. Your support will be a great motivation for me to continue bringing new models in the coming times.

Thank you so much for always being there and supporting me!

Recommend

  • Using forgeUI to run the model

  • Steps: ~20

  • Distilled CFG Scale: 3.5

  • CFG : 1

  • Sampling: Euler simple

  • Diffusion in Low Bits: Automatic

  • Vae: flux vae (must have)

  • Clip: clip_l.safetensors and t5xxl_fp8_e4m3fn.safetensors (must have)

________________________________________________________

StyleMix_1(fp8)

This is a Flux Dev FP8 version with a size of 11GB.

Recommend

  • Steps: 20

  • Distilled CFG Scale: 3.5

  • CFG : 1

  • Diffusion in Low Bits: float8-e4m3fn

  • Sampling: Euler simple

  • Nprompt: none

  • Vae: flux vae (must have)

  • Clip: clip_l.safetensors and t5xxl_fp8_e4m3fn.safetensors (must have)

________________________________________________________

Copax Timeless ver XPlus_1(GGUF Q8)/(GGUF Q4)


This is the GGUF version of model X Plus_1. This model has a file with the extension .gguf Use comfyUI to run this model.

The model's workflow you can download here: ComfyUI-GGUF simple workflow


!!Note: Version GGUF Q4 uses less vram but has a higher error rate than version GGUF Q8. You may have to regenerate a few times to get the desired results. If you want better quality, you can use GGUF Q8 or other versions.

Thank you pkmngotrnr, for helping me convert the model to the gguf version and providing me with usage instructions.

_____________________

Copax Timeless ver XPlus_1(NF4-LowVram)

This is the Flux Dev NF4 version of model X Plus_1


This version has a capacity of 6gb and does not include Clip and VAE, it uses less Vram.

If you want better quality, you can use version XPlus_1

Recommend

  • Using forgeUI to run the model (must have) ( Currently comfyUI node NF4loader cannot run this model)

  • Steps: ~20

  • Distilled CFG Scale: 3.5

  • CFG : 1

  • Sampling: Euler simple

  • Diffusion in Low Bits: bnb-nf4

  • Vae: flux vae (must have)

  • Clip: clip_l.safetensors and t5xxl_fp8_e4m3fn.safetensors (must have)

_________________________________________________________________

Copax Timeless ver XPlus_1

Hello everyone,

  • This model extends the capability to generate NSFW content for flux1-dev-fp8.
    This version focuses on improving Female Nudity, sex poses do not work well in the current version.

  • It's the first model I've created using Flux, and I recognize that there are still areas for improvement.

  • I plan to continue developing and enhancing this model in the future. Training a Flux model is very costly, so I've launched an Early Access option in the hopes of gaining support to fund the development of future versions.

If you enjoy the work I’m doing and want to see more updates in the future, please consider supporting me by leaving a like, review, tip, or making a donation. Your support will be a great motivation for me to continue bringing new models in the coming times.

Thank you so much for always being there and supporting me!

Recommend

  • Steps: 20

  • Distilled CFG Scale: 3.5

  • CFG : 1

  • Diffusion in Low Bits: float8-e4m3fn

  • Sampling: Euler simple

  • Nprompt: none

________________________

Ver XPlus_Unreal

This version created on Flux schnell can operate with only 4 steps, improving image creation speed. However, if you want to create more accurate images with prompts, you should not use this version. Use XPlus_1 version for better quality.

Recommend

  • Steps: 4

  • Distilled CFG Scale: 3.5

  • CFG : 1

  • Sampling: Euler simple

  • Diffusion in Low Bits: float8-e4m3fn

  • Vae: flux vae (must have)

  • Clip: clip_l.safetensors and t5xxl_fp8_e4m3fn.safetensors (must have)

____________________________

Copax Timeless ver XIII

Hello everyone,

I am thrilled to share with you the latest update, Copax Timeless XL ver XIII! This version brings several changes compared to the previous updates:

  • Character Appearance: Refreshed and enhanced to be more realistic and vivid.

  • Skin Details: Adjusted skin details, from texture to how light reflects off it.

  • Background: The scenery has become richer and more detailed.

  • Lighting Colors: The lighting effects have been fine-tuned to make your character blend seamlessly with the surrounding environment.

  • Hand and Costume Details: Improved hand details, with added data for costumes.

  • Style and Materials: Fine-tuned details for materials and styles.

If you enjoy the work I’m doing and want to see more updates in the future, please consider supporting me by leaving a like, review, tip, or making a donation. Your support will be a great motivation for me to continue bringing new models in the coming times.

Thank you so much for always being there and supporting me!

Recommend

  • Steps :25~50

  • CFG : 5~7

  • You can also use hiresfix or After Detailer

  • Sampling :

    • Restart

    • Euler a or DPM 2M ++ SDE Karass

  • Nprompt:

    (worst quality, low quality, illustration, 3d, 2d), open mouth, tooth,ugly face, old face, long neck,

______________________________________________

Copax TimeLess V12

Adjust skin gloss, skin details, improve hand and tooth details, add character appearance data, material details, background..etc..

Recommend

  • Steps :30~50 (Testing with step 50 and DPM 3M++ SDE Exponential gives quite good results)

  • CFG : 7

  • You can also use hiresfix or After Detailer

  • Sampling : DPM 3M++ SDE Exponential , DPM 3M++ SDE Karass, Euler a or DPM 2M ++ SDE Karass

  • Nprompt:

    (worst quality, low quality, illustration, 3d, 2d), open mouth, tooth,ugly face, old face, long neck,

________________________________________________

Copax TimeLess Version V11-lightning

The model is combined with the lightning model to get good image quality at a low number of steps

model work at step 8 and CFG 1. You can increase the number of steps to about 13 if you want finer detail. You can also try other settings but some of them may not work well

Recommend

  • Steps : 8~15

  • CFG : 1

  • Dimension: 768x1024, 1024x1024 (error may occur when using 576x1024 with adetailer)

  • You can also using After Detailer for face

  • Don't using Hires fix or setting denoising strength value at 0 when using hires fix. May affect the color of the photo

  • Sampling :DPM 2M ++ SDE Karass

  • Nprompt:

    (worst quality, low quality, illustration, 3d, 2d), open mouth, tooth,ugly face, old face, long neck,

______________________________________

Copax TimeLess Version V11

Adjust details, blur, noise. Additional data on astronomy, beach, alien, background, character, pose, effect,.. etc

Recommend

  • Steps : 30~60

  • CFG : 7

  • You can also use hiresfix or After Detailer

  • Sampling : DPM 3M++ SDE Exponential , DPM 3M++ SDE Karass, Euler a or DPM 2M ++ SDE Karass

  • Nprompt:

    (worst quality, low quality, illustration, 3d, 2d), open mouth, tooth,ugly face, old face, long neck,

______________________________________

Copax TimeLess Version V10

Adjust skin details, add more dark to images. Additional data on scenery, character appearance, pose, costumes, etc

Recommend

  • Steps : 25~50

  • CFG : 7

  • You can also use hiresfix or After Detailer

  • Sampling : DPM 3M++ SDE Exponential , DPM 3M++ SDE Karass, Euler a or DPM 2M ++ SDE Karass

  • Nprompt:

    (worst quality, low quality, illustration, 3d, 2d), open mouth, tooth,ugly face, old face, long neck,

___________________________________

https://www.sydigi.com

My email: [email protected]

Community

I have just completed the community pages so the content will be updated in the near future

________________________________________________________________________

Every dollar helps out to fund the GPU costs of training the model. Consider being a patron on Patreon to access stuff early. ❤️

If your computer isn't powerful enough or you don't want to spend time setting it up, you can generate images directly from this model on the online platform that supports GPU and unlimit generate images Mage.Space

_____________________________

Collection of sample images for models: Collection sample Images

Available models on Mage:

Copax Anime XL, Copax Timeless XL, Copax Timeless XL V7, Copax Timeless XL V8,

_______________________________

Copax TimeLess Version V9

This update significantly enhances the dataset, including detailed improvements in character appearances, environmental settings, specific details, styles, lighting effects, and color schemes.

Recommend

  • Steps : 25~50

  • CFG : 7

  • You can also use hiresfix or After Detailer

  • Sampling : DPM 3M++ SDE Exponential , DPM 3M++ SDE Karass, Euler a or DPM 2M ++ SDE Karass

  • Nprompt:

    (worst quality, low quality, illustration, 3d, 2d), open mouth, tooth,ugly face, old face, long neck,

_______________________________

_________________________________________________________

Copax TimeLess Version V8

Adding data related to character appearances, facial details, animals, natural landscapes, effects, colors, and more...

Recommend

  • Steps : 40~60

  • CFG : 7~10

  • You can also use hiresfix or After Detailer

  • Sampling : DPM 3M++ SDE Karass, Euler a or DPM 2M ++ SDE Karass

  • Nprompt:

    (worst quality, low quality, illustration, 3d, 2d), open mouth, tooth,ugly face, old face, long neck,

_____________________________

Copax TimeLess Version V7

Improved skin details, facial features.

_____

Recommend

  • Steps : 40~60

  • CFG : 5-7

  • You can also use hiresfix or After Detailer

  • Sampling : Euler a or DPM ++ SDE Karass

Negative Prompt

painting, drawing, sketch, cartoon, anime, manga, render, CG, 3d, watermark, signature, label, long neck, misaligned eyes, teeth

_______________________

====================================
About focused on developing diversity in styles


I'm delighted because with every update of this model, it receives a multitude of reviews from people. The reviewed images showcase a wide variety of styles.

It's fascinating and brings about pleasant surprises, not limited to any specific style.

My goal is to provide a model that allows people to convey and express their ideas through images.


It's because of this that this model is focused on developing diversity in styles rather than concentrating on a specific one.


It's also because of the complexity of various types of data that the model won't excel in any specific style but aims for balance.

==================================================

About V6

The demo images were created using the Euler A and a low step value of 28. At this step, the images exhibit a blur effect, artistic style, and do not display detailed skin features. If you wish to have more details, you can render them at a higher step, around 50 to 60.


Version 6 is now stable and can run smoothly at a lower step value of 28, which is lower than other versions that require steps ranging from 40 to 60.

Copax TimeLess Version V6

Add character data, facial details, eyes, style, reduce frame error rate, backgrounds, effects, reduce hand detail error rate when generating images and more...

Recommend

Steps: ~28-60,

CFG scale: ~4-10

Sampler: euler a / DPM++ 2M SDE Karras

Adetail for face

N prompt:

(worst quality, low quality, illustration, 3d, 2d), open mouth, tooth,ugly face, old face, long neck,

==============

Version 5 marks the beginning of the phase where I'm adding more data about materials, mechanical details, physical effects, and environments. I will continue to enhance the data with upcoming updates.

The issues identified in Version 4 include:

  1. Issues with hand and feet drawing, character neck proportions, frame ratios,

  2. Causing difficulty in harmonizing characters with the background at times.

I had a predefined data training plan, these concerns will be addressed in Version 6.

=========

Update N prompt:

Thank you, sevenof9247 for sharing. Your N prompts are truly effective in rendering realistic images. Link article author:

https://civitai.com/articles/1826/sdxl-prompts

N prompt:

painting, drawing, sketch, cartoon, anime, manga, render, CG, 3d, watermark, signature, label

=================

About run model on website

The demo images I used ComfyUI. You just need to download the image and drag it into the ComfyUI interface, and it will display the workflow, prompt, and settings for the image.

Civitai remove the prompt meta and change to a .jpg file with images rendered by ComfyUI.

Model Name Change

I would like to express my gratitude to all of you for using the model, providing likes, reviews, and supporting me throughout this journey.

I have planned to train the model with each update version.

My intention is to gradually enhance the model's capabilities with additional data in each version.

This includes adding skin details, expanding physical features, incorporating facial details for various regions like Europe, Asia, and Africa, as well as catering to diverse age and gender variations. Furthermore, I will focus on material details, effects, lighting adjustments, and more.

The process of training for realism demands high-quality images and extensive training steps. This requires substantial effort and preparation time. The model's improvement in realism will come about through several rounds of training.

However, the choice of the name "Realistic" unintentionally set exceedingly high expectations for a highly realistic model, even though the model is still in its developmental phase. This situation has created significant pressure on me to rapidly enhance the model's realism.

Therefore, I have decided to rename the model to "TimeLessXL," inspired by my Model 1.5. This new name allows me the freedom to explore various styles, whether they be 2.5D, bright tones, dark tones, or 3D, without being bound by the necessity for realism.

I will continue to provide updates and enhance the model with realistic details and effects. However, I will no longer feel the urgency to rush the training process or solely focus on achieving absolute realism.

I sincerely thank you for taking the time to read this announcement, and I hope for your continued support in the times ahead.

=======================================================

Copax TimeLess Version V5

Adding data about materials, rust effects, mechanical details, people, scenes, and more..

Recommend

Steps: ~40-60,

CFG scale: ~4-10

Sampler: euler a / DPM++ 2M SDE Karras

Adetail for face

N prompt:

painting, drawing, sketch, cartoon, anime, manga, render, CG, 3d, watermark, signature, label, long neck,

=======================================================

Copax TimeLess Version V4.6

Adjust character details, fine-tune lighting, and background. Enhance the contrast between the person and the background to make the subject stand out more.

Recommend

Steps: ~40-60,

CFG scale: ~4-10

Sampler: euler a / DPM++ 2M SDE Karras

Adetail for face

N prompt:

painting, drawing, sketch, cartoon, anime, manga, render, CG, 3d, watermark, signature, label

=====================================================

Copax TimeLess Version V4.4

Add more skin details, middle-aged, elderly, Asian male, European male, adjust colors and brightness...etc.

Recommend

Steps: ~40-60,

CFG scale: ~4-10

Sampler: euler a / DPM++ 2M SDE Karras

Adetail for face

N prompt:

Thank you, sevenof9247 for sharing. Your N prompts are truly effective in rendering realistic images. Link article author:

https://civitai.com/articles/1826/sdxl-prompts

N prompt:

painting, drawing, sketch, cartoon, anime, manga, render, CG, 3d, watermark, signature, label

=======================================================

Copax TimeLess Version V4.2

Adding more data about people, facial details, materials, etc.

Recommend

Steps: ~40-60,

CFG scale: ~4-10

Sampler: euler a / DPM++ 2M SDE Karras

Adetail for face

N prompt:

(worst quality:1.6), (low quality:1.6), (normal quality:1.6), lowres, bad anatomy, bad hands, vaginas in breasts, ((monochrome)), ((grayscale)), collapsed eyeshadow, multiple eyebrow, (cropped), oversaturated, extra limb, missing limbs, deformed hands, long neck, long body, imperfect, (bad hands), signature, watermark, username, artist name, conjoined fingers, deformed fingers, ugly eyes, imperfect eyes, skewed eyes, unnatural face, unnatural body, error, painting by bad-artist

=======================================================

Copax TimeLess Version V4

Version 4 enhances facial features, hair, physique, skin details, and more.

Recommend

Steps: ~60,

CFG scale: ~6-10

Sampler: DPM++ 2M SDE Karras

Adetail for face

N prompt:

(worst quality:1.6), (low quality:1.6), (normal quality:1.6), lowres, bad anatomy, bad hands, vaginas in breasts, ((monochrome)), ((grayscale)), collapsed eyeshadow, multiple eyebrow, (cropped), oversaturated, extra limb, missing limbs, deformed hands, long neck, long body, imperfect, (bad hands), signature, watermark, username, artist name, conjoined fingers, deformed fingers, ugly eyes, imperfect eyes, skewed eyes, unnatural face, unnatural body, error, painting by bad-artist

Link file images demo:
I use random prompts generated by the SDXL Prompt Styler, so there won't be any meta prompts in the images.

=======================================================

Copax Realistic Version Colorful V2


Version 2 introduces additional details for physical appearances, facial features, etc. Character images and color ranges are now more distinct and clearly separated from each other.

Link file images demo: I will provide the download links for the demo images tomorrow.

=========================================================

Copax Realistic Version Colorful

The colorful version includes additional effects and colors.

This version does not upgrade any realistic details. Use version 3 if you don't need color effects.

==================================================

Copax Realistic Version 3 - SDXL1.0

This update adds more pose, detail, style..vv


The demo images I used ComfyUI. You just need to download the image and drag it into the ComfyUI interface, and it will display the workflow, prompt, and settings for the image.

Link file images demo: https://huggingface.co/datasets/Copax/demo_model/tree/main

Link file images demo part 2: https://civitai.com/models/123333?modelVersionId=134451

Link: Workflow ComfyUI for Copax Realistic XL V3

Civitai remove all metadata and change to .jpg file.
Therefore, I have packaged the demo image files and uploaded them to Huggingface so that everyone can download the original .png files.

My computer configuration can only run ComfyUI, and I have to rent a server for the entire model training process.

Recommend

Steps: ~60,

CFG scale: ~3-10

Sampler: DPM++ 2M SDE Karras

Adetail for face

Sample prompt:

dreamlikeart, 3d render of futuristic military cyborg, cyberpunk, wolf, warframe, stealth, armored, neon lights, background explosion, character design, hard surface, smooth, detailed face, highly detailed, intricate details, symmetrical, volumetric lighting, ambient light, real-time, vfx, digital 3d, uhd, hdr

N prompt:

(worst quality:1.5), (low quality:1.5), (normal quality:1.5), lowres, bad anatomy, bad hands, multiple eyebrow, (cropped), extra limb, missing limbs, deformed hands, long neck, long body, (bad hands), signature, username, artist name, conjoined fingers, deformed fingers, ugly eyes, imperfect eyes, skewed eyes, unnatural face, unnatural body, error, painting by bad-artist, (worst quality, low quality,nsfw,nipple, pussy:1.5)

========================================================

Copax Realistic XL Version 2.2 - SDXL1.0

This update adds more skin details.

Recommend

Steps: 70,

CFG scale: ~6-10

Sampler: Euler a, DPM++ 2M SDE Karras

Hires upscaler: 4x-UltraSharp,

=======================================================

Copax Realistic XL Version 2 - SDXL1.0

Recommend

Steps: 50,

Sampler: DPM++ 2M SDE Karras, CFG scale: 4

Denoising strength: 0.4,

Hires upscaler: 4x-UltraSharp,

===========================================

Copax Realistic XL Version 1 - SDXL1.0

Finetuned over SDXL1.0

RECOMMEND

Steps: 50,

Sampler: Euler a,

CFG scale: 3~7

VAE: no vae / vae SDXL

N prompt:

2d art, 3d art, ((illustration)), anime, cartoon, bad_pictures, bad-artist, EasyNegative,(worst quality:1.6), (low quality:1.6), (normal quality:1.6), lowres, bad anatomy, bad hands, vaginas in breasts, ((monochrome)), ((grayscale)), collapsed eyeshadow, multiple eyebrow, (cropped), oversaturated, extra limb, missing limbs, deformed hands, long neck, long body, imperfect, (bad hands), signature, watermark, username, artist name, conjoined fingers, deformed fingers, ugly eyes, imperfect eyes, skewed eyes, unnatural face, unnatural body, error, bad image, bad photo