Electroverted's Everything Everywhere All At Once txt2img ComfyUI Workflow版本SDXL v1.1 (ID: 234016)

"I got bored one day and put everything on a bagel."

Everything Everywhere All At Once (formerly E-Z All-Inclusive) is an encompassing txt2img workflow. Don't worry, it's organic – no custom nodes needed. A guide is included, and if you're new to this, download the latest version and then click on older versions for visual examples of the interface. This is the workflow I use for txt2img, and I love it.

EEAAO has toggles to switch between universes:
- random seed output
- CFG variations of one seed
- four different upscale variations and a quick upscale refiner
- SDXL or SD 1.5
- 3 LoRAs and 1 SDXL ControlNet
- Sucked ♫ into ♫ a bagel ♫

I split the positive and negative prompts to utilize SDXL's CLIP G, which is far more dynamic than the old CLIP L. I still use CLIP L to boost certain words in a prompt, like camera angles and lighting, but the more you put in there, the less variety you'll get in outputs. If you're really struggling to get what you want, put the same prompt in both sections, but be advised that the CLIP L boost really limits some of the variety you get from CLIP G.

SDXL ControlNet models (required if you want to use ControlNet):

https://huggingface.co/stabilityai/control-lora/tree/main/control-LoRAs-rank256

Custom node manager (not required):

https://github.com/ltdrdata/ComfyUI-Manager

描述:

Added SD 1.5 CLIP support.

Fixed a CFG bug in the upscalers.

Tweaked some upscale settings.

Wrote some more notes.

训练词语: