CADS3 XL based on Transformer's diffusion model (Di-T)

AiARTiST SDXL commercial design and advertising material workflow

With the release of version SD v3, we can see the diffusion model of the new architecture DiT, which has more possibilities for picture composition

, although V3 has also been criticized due to issues with licenses and fragmented training sets, the new version's improvements in picture quality and composition cannot be denied.

Therefore, after half a month of "research" and "analysis", CADS has launched the 3.0 version of the XL commercial design model, aiming to bring closer to the stable diffusion practical model of the Di-T era in terms of picture quality, material, composition ability, and understanding ability.

Workflow based on the sensitive OMOST architecture and omost-llama-3-8b-4bits large language model

Omost project address:

https://github.com/lllyasviel/Omost

Omost ComfyUI component author project address:

https://github.com/huchenlei/ComfyUI_omost

Also, you can pay attention to the LLM automation project (support OMOST API push):

https://github.com/heshengtao/comfyui_LLM_party

Model and plugin file download cloud disk:

Link: https://pan.baidu.com/s/1mb4MPxx1mdIR2YCh1OWXzA?pwd=AIGC

Extraction code: AIGC

OMOST Di-T

Author of ControlNet's new project, also involving LLM large model and Agent.

Of course, still related to AI painting: solving the automation needs of people's propositions or compositions.

Now, with just a concrete or abstract ["theme"], the Agent will start "creative thinking" on its own.

Although the creative ability is currently a bit erratic, and the resulting images are sometimes awkward,

this is ultimately a new direction for assisting (gradually replacing) human creativity.

Notice:

The text prompts of OMOST are not describing the picture like the common prompts of SD,

but provide a "theme" or "basic idea", for example:

Love and Peace  /  A comedy movie scene /  Wealthy women surrounded and so on

LLM natural language model will formulate and plan the content of the image according to our requirements.

If you are not satisfied with the output, you can request precise modifications to the generated image content via the modify command in the lower right corner.

Other support groups:

Portrait repair FaceRestore -Github official website

https://github.com/mav-rik/facerestore_cf

----------------------------------------------------------------------

AiARTiST

SDXL's SD3 simulator LoRA is online:

SD3 simulator-XL-V3-Simulator-LoRA

https://www.liblib.art/modelinfo/7449642bdf39443aa66341837025c690

----------------------------------------------------------------------

Lib members can enjoy online accelerated generation, convenient matching with CADS series LoRA.

Model list: https://www.liblib.art/search?keyword=AiARTiST

Free model ratio 55%, open-source model ratio 95%

Member model ratio 40%, closed source model ratio 5%

Commercial models:

AiARTiST-CADS2.0 XL Commercial Advertising Assistance System (Enterprise Customization)

https://www.liblib.art/modelinfo/a56ebacdba7d4e30b97bb124bc3fc28f

AiARTiST-CADS3.0 XL Di-T Commercial Design Assistance System (currently only available for local customization deployment)

https://afdian.net/a/AiARTiST-lib?tab=shop

--------------------------------------------------------------------

For testing questions, please leave a message, for business cooperation, visit the personal homepage: +V Zyuan980

AiARTiST | Metaverse China·Shandong Digital Human/AiGC/Metaspace direction

Being a good tool to serve artists, contact information: https://afdian.net/a/AiARTiST-lib