动漫Comfyroll Template Workflows

此模型的原作者为「@Akatsuzi」,我们致力于推动原创模型的分享和交流,因此转载了该模型、供大家非商用地交流、学习使用。请使用该模型的用户,遵守原作者所声明的使用许可。

如果您是该模型的原作者,请与我们联系,我们期待您入驻、并将第一时间把模型转移至您的账号中。如您不希望在 分享该模型,我们也将遵循您的意愿在第一时间下架您的模型。

我们尊重每一位模型原创作者,更期待与每一位模型作者共同成长!

声明:若该转载模型引发知识产权纠纷或其他侵权行为,我们将立即下架模型,并不会向原作者追责。

This collection of ComfyUI workflow templates is designed for use with the Comfyroll checkpoint, but can be used with any checkpoint model.

https://civitai.com/models/58712/comfyroll

The initial collection comprises of the following:

  1. A simple workflow with a variable semi-realistic style prompt

  2. A simple workflow for creating anime style illustrations

  3. An intermediate workflow for creating more complex stylized anime style illustrations

  4. A simple workflow for creating simple stylized illustrations

  5. An intermediate workflow for creating complex stylized illustrations

  6. An advanced workflow with a large variety of options

It is planned to add more templates to the collection over time.

Img2Img noise masks:

The download includes a folder containing noise masks sourced from Wiki Commons. These are used in the workflows. The masks can be used to enhance the pseudo-HDR effect and also to add detail and style in the output.

There are many other types of noise, and many royalty free sources where these can be found. For example, textures and halftone images can make good noise masks.

Style masks:

The download also contains a folder for style masks. Style masks are masks that impact the composition and coloring used in the output. They can be made from textures, logos, and many other image types. Here are a couple of examples from royalty free sources.

Company logos https://commons.wikimedia.org/wiki/Category:Logos_of_companies_by_industry

Alphabetic characters

https://commons.wikimedia.org/wiki/Category:Single_letter_icons

Installation

You can save the workflow templates anywhere.

The Custom Nodes in the download should be saved in your \ComfyUI\custom_nodes folder.

Download and install WAS Node Suite from https://civitai.com/models/20793/was-node-suite-comfyui

Troubleshooting

If you have an older version then the model name may need updating or an image mask may need connecting. You can update these in the Load Checkpoint and Load Image nodes.

Sample noise mask images can be found in the download.

If you have other errors then try re-downloading and check to see if there is a more recent version of the template in the download.

Workflow templates:

General notes on templates:

  • Nothing about the templates should be regarded as unchangeable. Please feel to modify them as much as you want. New styles can be discovered through experimention.

  • Clip Skip -1 provides a slightly less realistic cutesy style, but is also good for some illustration styles

  • Clip Skip -2 provides a semi-realistic style that also combines well with illustration styles

  • Increasing the sampling steps provides a smoother image. Lower sampling steps produce a more contrasty manga like style.

  • All of the templates require custom switch nodes provided in the download. These are colored red in the templates and can be set to either 1 or 2 to control the flow through the workflow.

  • Most of the templates also require custom nodes from WAS Node Suite

    • https://civitai.com/models/20793/was-node-suite-comfyui

    • this is highly recommended

Template 1 - Simple variable prompt

  • this template can be used to produce semi-realistic or anime style images

    • use Comfyroll for semi-realistic style

    • use ComfyrollAnime for anime style

  • it uses a variable prompt that can be used for experimentation and discovery

  • the style can easily be changed by modifying the style line

    • e.g. semi-realistic, HDR, 2.5D,

  • the color palette can easily be changed by modifying or removing the color criteria line

    • e.g. colorized, black + red + green + yellow limited color palette,

  • additional variables can be easily added, e.g. {summer|winter|autumn|spring|}

  • existing variables can be easily modified be adding or deleting items within the curly brackets

Template 2 - Simple anime style template

  • this template produces anime style images

  • the style can easily be changed by modifying the style line

    • e.g. concept drawing, illustration, manga, anime line-art style,

  • the color palette can easily be changed by modifying or removing the color criteria line

    • e.g. colorized, pink + light-blue + white + purple limited color palette,

  • additional variables can be easily added, e.g. {swimsuit|bikini|sportswear|}

  • existing variables can be easily modified be adding or deleting items within the curly brackets

  • two sample images are provided using this template

Template 3 - Intermediate anime-style stylized illustrations

  • this template produces more complex anime-style stylized images

  • it works best with the ComfyrollAnime model

  • it is similar to the simple anime-style template but with the following additional features

    • a ControlNet to apply a noise mask after the hi-res upscale

    • padding on the image mask

    • four manual switch points

    • blurring on the image output

  • additional features are planned for later versions

  • this is similar to using a ControlNet on the Img2Img tab in Auto1111, but the ControlNet is applied after the Hires fix

Template 4 - Simple semi-realistic stylized illustrations

  • this template produces simple semi-realistic stylized images

  • the style can easily be changed by modifying the style line

    • detailed concept drawing, illustration, line-art style, watercolor,

  • the color palette can easily be changed by modifying or removing the color criteria line

    • colorized, red + green + blue + yellow limited color palette,

  • the coloring and complexity of the image mask will impact the final output

  • logos and textures make good image masks for stylized pictures

  • halftone images also make great masks

  • panelled masks can be used to create panelled output

  • recommended denoising between 0.85 and 0.95

Template 5 - Intermediate semi-realistic stylized illustrations

  • this template produces more complex stylized images

  • it is similar to the simple template but with the following additional features

    • a ControlNet to apply a noise mask after the hi-res upscale

    • padding on the image mask

    • optional switch between latent upscale and hires fix

    • blurring on the image output

  • this is similar to using a ControlNet on the Img2Img tab in Auto1111, but the ControlNet is applied after the Hires fix

Template 6 - Advanced Workflow

  • the template is packed with different options and may take some time to get familiar with, but will be worth the effort

  • this is the workflow that i use on a daily basis to create most of my AI illustrations

  • it includes 11 manual switching points for controlling the flow

  • each optional component can be bypassed using the switches

  • this template is similar to template 5 but with the following additional features

    • dual prompts at different stages in the workflow

    • dual ControlNets at different stages in the workflow

    • dual loras

    • batch process for loading Images or ControlNets

    • more switch points

    • image preview (pre-hires)

  • this is similar to using batch Img2Img and multi-ControlNet in Auto1111, but with lots of additional options

  • set the Load Image Batch to single-image when it is not being used

  • the default setting on all switches is 1 (Off)

Sources:

White Noise Mask https://commons.wikimedia.org/wiki/File:White.noise.b.w.png

Pink Noise Mask https://commons.wikimedia.org/wiki/File:Pink.noise.b.w.png

Dissolve Noise Mask https://commons.wikimedia.org/wiki/File:512x512_Dissolve_Noise_Texture.png

Perlin Noise Mask

https://commons.wikimedia.org/wiki/Category:Perlin_noise#/media/File:Perlin_noise_example.png

Comfyroll Custom Nodes by https://civitai.com/user/RockOfFire6

Change History:

v21

Advanced Template

  • fixed ControlNet conditioning

  • added LoRA for 2nd prompt

Placeholder post

AI模型详情

模型类型:Other
基础算法:基础算法 v1.5
触发词:
文件格式: zip
最近更新:2024-02-10 01:27
创作相关
允许下载:
严禁转载:
会员专享下载:
作者原创:
转载来源: liblib
转载说明: 该模型及相关图片信息可能涉及来源于其他开源社区或由用户自行上传发布,非原创作者,如需商业使用请联系原作者。
郑重声明:本模型仅用于练习使用模型,如需商业请与原作者联系获取版权许可等,互联网鼓励用户分享作品,并坚决保护创作者的版权信息,模型、图片、素材资源的版权均为原创作者所有。
免责声明: 您基于本模型服务生成的文本内容由您自行维护并对其独立判断后使用,您需在遵守适用法律法规和本服务条款的条件下使用,基于本服务生成内容产生的任何知识产权问题由您自行处理,您需对生成内容负责,本网站对由此造成的任何损失不负责任。除法律法规另有规定外。
版权说明: 本网站部分素材、模型、图片等信息基于用户自行上传和分享,此资源来源于网络,如涉及版权问题,请及时与本网站联系,如需删除收到通知响应后,会在处理后下架删除。

讨论列表 共有 0 条讨论

暂无评论信息...
为用户提供AI模型下载、AI模型训练、AI模型学习、AI模型教程。
AI模型分享社区,基于Stable Diffusion潜在扩散的机器学习模型。