smdb NoobAI - vpred workflow版本v1.1 (ID: 1300995)

ComfyUI workflow to generate good quality images with base NoobAI vpred model and a refiner.

Known issues

For unknown reason batch size 2 generates worse quality images than batch size 1. I don't know if this is limitation of my low VRAM GPU (I use old RTX 3060 8Gb) but anyway try to start with batch size 1 and than compare the results with batch size 2,

Resources used to generate samples gallery

1. Base model: NoobAI vpred 1.0. I like to use vanilla NoobAI for maximum flexibility, artists styles and Lora support.

2. Refiner model: NoobAI Cyberfix-V2 EPS. Universal all-rounder, very close to base NoobAI. NoobAI Cyberfix-V2 vpred has strongest artist styles and slightly better anatomy but generates darker images (can be color-corrected in the workflow). For 2D art I also like ZUKI cute ILL but it changes character faces compared to base.

3. Upscaler: 4x-NomosUni_span_multijpg fast upscaler model.

Workflow bookmarks /hotkeys

• Keyboard hotkey (1): Models select

• Keyboard hotkey (2): Prompt

• Keyboard hotkey (3): txt2img Settings

• Keyboard hotkey (4): Upscale settings

• Keyboard hotkey (0): Output Images

I use new native reroutes feature, enable it in ComfyUI Settings -> Lite graph.

---

If you have any questions leave the comments below.

描述:

• Adjusted default sampler settings. Should works better both with 2D and 2.5D.

• Support prompt scheduling in negative prompt.

• Adjusted default positive and negative prompts.

训练词语: