Sign In

Retro Graphic Design TI XL

24
176
152
7
Updated: Jul 11, 2024
style
Verified:
SafeTensor
Type
Embedding
Stats
176
152
6
Reviews
Published
Jul 11, 2024
Base Model
SDXL 1.0
Training
Steps: 2,000
Epochs: 1
Usage Tips
Clip Skip: 1
Trigger Words
cs-gdblk
Hash
AutoV2
B9A579F0AF
default creator card background decoration
ODOR Badge
chromesun's Avatar
chromesun

2024-JUL-11

A TI (Textual Inversion) embedding to make retro graphic design images with a 20th century vibe. Think vinyl album covers, Omni magazine, CMYK process colors. At least, that’s what I’m thinking :-) Perhaps this is a recognised (sub-)genre? I had a look through CivitAI and asked the google, but I couldn’t find a match for what I was imagining... so I made my own.

Anyone got a better label?

This TI uses 8 vectors (i.e. 8 tokens in your prompt) and is 33KB in size. Weighting works (tested in range 0.5 <---> 1.2), but I usually just move the trigger right/left to weaken/strengthen respectively. I did not use any "in the style of” artists or art movements in the training images.

The trigger word is cs-gdblk, but you can change that simply by renaming the .safetensors file. If you do, try not to use a real word that SDXL already knows!

Behaviour:

* Graphic style rather than photorealism.

* Reduced color palettes, e.g. CMYK, ROYGBIV, duo/tri/quadtone.

* Can produce detail if you ask for it, but defaults to simpler shapes and block colour for short prompts.

* Vibrant output in most checkpoints I tested.

* More darker/night backgrounds than lighter. By design; this TI is gdblk=Graphic Design Black. I’ve been trying to make a complementary Graphic Design White TI for lighter/day backgrounds but without success so far.

Testing and showcase images done in Forge version: f0.0.17v1.8.0rc-latest-276-g29be1da7.

Sampler: testing worked well with...

DPM++ 2M Karras 25<--->50 steps, generally I used 30 or 40 in testing

CFG scale: 5<--->10, generally I used 6 or 7 in testing

Mostly used Hires.fix at either 1.25 or 1.5 to increase the detail a bit.

Hires.fix steps 15<--->25, 4x_NMKD_Siax_200k or your favorite, denoising 0.4

I didn't use any other adjusters/controlnet/i2i/post-processing for the showcase images.

During testing I kept the negative prompt as simple as possible, e.g.:

closed eyes,signature,logo

Checkpoint models that worked well when I tested this TI:

Clarity XL

https://civitai.com/models/471585/clarity-xl

Magie_Noire v4

https://civitai.com/models/505656?modelVersionId=612138

Midnight v5

https://civitai.com/models/496960/midnight

BaZaar v0.20_UD

https://civitai.com/models/450025

ArtOlives v1

https://civitai.com/models/369548

Crystal Clear XL

https://civitai.com/models/122822/crystal-clear-xl

ZavyChromaXL v8

https://civitai.com/models/119229?modelVersionId=563988

[PUBLISHEDTOCIVITAIONLY]