santa hat
deerdeer nosedeer glow
Sign In

Doomer Boomer

744
3.7k
139
Verified:
SafeTensor
Type
Checkpoint Trained
Stats
3,701
Reviews
Published
Jul 29, 2023
Base Model
SD 1.5
Usage Tips
Clip Skip: 1
Trigger Words
by Boris Vallejo
by Brom
by Frank Frazetta
by Hajime Sorayama
by Kelly Freas
by Luis Royo
by Milo Manara
by Olivia De Berardinis
by Philippe Druillet
by Simon Bisley
Hash
AutoV2
798744A463

Doomer Boomer

A 1980s - 1990s fantasy/sci-fi RPG art model.


If you're family with my other models you might be asking yourself, haven't I seen this before?

So here's a quick breakdown of the differences between seemingly similar models :

Electric Eden : My first attempt at creating a 80s - 90s RPG style art model. Using a small sample model (trained on a small portion of the full dataset) and then merged with a few other models to makeup for being undertrained.

Another Damn Art Model (ADAM) : Further trained the model from Electric Eden and merged different epochs to bring out elements from each epoch that I liked. Still trained on a sample set of the main dataset.

Doomer Boomer : Unhappy with some of the results in Another Damn Art Model (ADAM); I reevaluated my dataset. Completely starting over and creating a new training set. This time, using 768x768 images instead of 512x512. I also focused on structuring my training set so that's it's more organized, using data processing techniques to clean the directories and ensure that there wasn't any artists images being wrongfully attributed to another.

I also increased the number of quality images from each artists. At the coast of reducing the number of artists down to exactly 10 artists (you can see a list of them bellow). Once again, this model is still not using the full (new) dataset. This was trained on yet another sample dataset for testing purposes. Pleased with the results I was seeing, I decided to release it here.


Questions/Feedback/Updates?

Visit my thread on the Unstable Diffusion Discord


Artists :

  • Boris Vallejo

  • Brom

  • Frank Frazetta

  • Frank Kelly Freas (or just Kelly Freas, for tagging)

  • Hajime Sorayama

  • Luis Royo

  • Milo Manara

  • Olivia De Berardinis

  • Philippe Druillet

  • Simon Bisley

Simply add by in front of the artists name in the prompt to trigger their style.

For example, by Simon Bisley

If the style isn't strong enough. Use attention/emphasis; (by Simon Bisley:1.2)


VAE Required - Download it here

(the full model will have it's own fine-tuned VAE).

Install :

stable-diffusion-webui -> models -> VAE

In Webui :

Under Settings -> Stable Diffusion -> SD VAE select the downloaded VAE and uncheck Ignore selected VAE for stable diffusion checkpoints that have their own .vae.pt next to them


Settings

Quick Settings :

  • Settings -> User interface -> Quicksettings list

  • sd_model_checkpoints sd_vae CLIP_stop_at_last_layers s_chrun always_discard_next_to_last_sigma

ETA Noise Seed Delta :

  • Settings -> Sampler parameters -> Eta noise seed delta

  • 31337

Sampler Method :

  • DPM++ SDE Karras

    • Sampling steps = 25 - 40

    • CFG = 7 - 9

  • Euler a

    • Sampling steps = 20 - 30

    • CFG = 6 - 7.5

These are just recommendations. Feel free to experiment with other samplers and different parameters.


Hires Fix

Models :

Install location : stable-diffusion-webui -> models -> ESRGAN

Hires Fix Settings :

For all ESRGAN models (which the above are) I use the following settings :

  • Upscale by

    • 1.5 if resolution is > 512x768

    • 1.6 if resolution is <= 512x768

  • Denoising strength

    • 0.25 - 0.35

  • Hires steps

    • If sampling steps > 60 then,

    • Hires steps = half of sampling steps.

    • for example, sampling steps = 80, then Hires steps = 40.


ADetailer

Extension for webui that uses ultralytics to 'fix' faces/hands/body after generation.

Download here - https://github.com/Bing-su/adetailer

Read the information in the GitHub repo linked above for more info and installation guide.


Check out my other models :

SDXL

SD1.5

LoRA