# This is not an AI Art Podcast (Ep. 6)
## Intro
Welcome to episode six! This is your host, Doug Smith. This is Not An AI art podcast is a podcast about, well, AI ART – technology, community, and techniques. With a focus on stable diffusion, but all art tools are up for grabs, from the pencil on up, and including pay-to-play tools, like Midjourney. Less philosophy – more tire kicking. But if the philosophy gets in the way, we'll cover it.
But plenty of art theory!
Today we've got:
* Model madness model reviews: 2 loras, a textual inversion, a dreambooth and a model.
* "Bloods and crits": Art critique on 3 pieces
* Technique of the week: Using `reference_only` control net as well as region segmentation control net.
* My project update: so you can learn from my process
Available on:
* [Spotify](https://open.spotify.com/show/4RxBUvcx71dnOr1e1oYmvV)
* [iHeartRadio](https://www.iheart.com/podcast/269-this-is-not-an-ai-art-podc-112887791/)
* [Google Podcasts](https://podcasts.google.com/feed/aHR0cHM6Ly9hbmNob3IuZm0vcy9kZWY2YmQwOC9wb2RjYXN0L3Jzcw)
Show notes are always included and include all the visuals, prompts and technique examples, the format is intended to be so that you don't have to be looking at your screen -- but the show notes have all the imagery and prompts and details on the processes we look at.
## News update
Stable studio has been open sourced!
https://github.com/Stability-AI/StableStudio/blob/main/packages/stablestudio-plugin/README.md
Y'know how I've been talking about the automatic1111 community/ecosystem? This is a game changer.
Expect to see lots of action in the next few weeks, lots of ports. AFAIU there's kinda no "engine" in this yet, just a rolling chassis with the stability AI API ripped out, and someone's gotta add the backend back in it.
This should be fire.
## Model Madness
### Watercolor LoRA
* [From Reddit](https://www.reddit.com/r/StableDiffusion/comments/13g1ilm/watercolor_lora/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=share_button)
* [On Civitai](https://civitai.com/models/64560?modelVersionId=69190)
```
a watercolor painting of a 1920's flapper <lora:watercolorV1:0.0>
Negative prompt: bad_prompt_version2:0.8
Steps: 25, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: -1, Face restoration: CodeFormer, Size: 512x768, Model hash: 9aba26abdf, Model: deliberate_v2
```
The creator says it's still a WIP and that it may be over fit.
I put together an x/y plot. And I'd say generally you want to use this at a max of 0.5 strength, maybe even 0.3 or more is better. Getting stuff that's far too similar as the strength increases.
### Cassandre Artstyle (Dreambooth)
https://huggingface.co/jjraii/CassandreArtstyle
Based on: https://en.wikipedia.org/wiki/Cassandre
Rather effective. Likely worth a try. Could be cool to trim out the text and use them for control nets because they do come out with some good compositions.
```
a colorful poster of a lake house, cassandre style
Negative prompt: bad_prompt_version2:0.8
Steps: 25, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 2927782593, Face restoration: CodeFormer, Size: 512x768, Model hash: 3df7f74cd7, Model: Cassandre Artstyle V1
```
### Micro/Macro Textual inversion
* [Reddit thread](https://www.reddit.com/r/StableDiffusion/comments/13hlk2f/micro_and_macro_textual_inversion/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=share_button)
* [Micro TI on civitai](https://civitai.com/models/66142/micro-type)
* [Macro TI on civitai](https://civitai.com/models/66154/macro)
Micro has a good kind of model rail road model look.
```
action figure in a tropical cave, kkw-micro
Negative prompt: bad_prompt_version2:0.8
Steps: 25, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 3715793814, Face restoration: CodeFormer, Size: 768x512, Model hash: 0b914c246e, Model: analogMadness_v40
```
For macro, you still gotta help it, with just `cat in Brussels` (basically) I'd get a big cat every 5ish generations.
```
a (giant cat:1.2) walking through Grand Place in Brussels Belgium, night, kkw-macro, RAW photo, analog style
Negative prompt: bad_prompt_version2:0.8, (drawing, illustration, 3d render, cartoon:1.2)
Steps: 25, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 601809730, Face restoration: CodeFormer, Size: 768x512, Model hash: e3055c494a, Model: elegance_37
```
### Vector Art LoRA
* [on Reddit](https://www.reddit.com/r/StableDiffusion/comments/13esneo/new_vector_art_lora_on_civitai/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=share_button)
* [on Civitai](https://civitai.com/models/63556?modelVersionId=68115)
Looks good! I'd definitely see myself using this.
```
1920s flapper in the style of 0mib <lora:vectorart_0mib:1>
Negative prompt: bad_prompt_version2:0.8
Steps: 25, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 277164392, Face restoration: CodeFormer, Size: 512x768, Model hash: a60cfaa90d, Model: dreamshaper_5BakedVae
```
### RPG v4
Apparently [v5 is on the way, according to this reddit thread](https://www.reddit.com/r/StableDiffusion/comments/13hg75u/rpg_v50_early_result/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=share_button)
It's on [Civitai](https://civitai.com/models/1116/rpg)
This thing is ridiculous, and FUN.
There's a huge [PDF guide](https://huggingface.co/Anashel/rpg/resolve/main/RPG-V4-Model-Download/RPG-Guide-v4.pdf) -- the best I've seen to come with any model.
Has pointers for all kinds of stuff it's trained on, like character classes and clothing and armor types. Amazing! Wow I could just make characters all day.
```
full body, walking pose, slow motion, (rogue ranger:1.2) Swedish female wearing full body multiple stitched layer (fur armor:1.2),(white silk cloak), (fabric with intricate pattern:1.2), (insanely detailed, bloom:1.5), (highest quality, Alessandro Casagrande,Greg Rutkowski, Sally Mann, concept art, 4k), (analog:1.2), (high sharpness), (detailed pupils:1.1), (painting:1.1), (digital painting:1.1), detailed face and eyes, Masterpiece, best quality, (highly detailed photo:1.1), 8k, photorealistic, By jeremy mann, by sandra chevrier, by maciej kuciara, sharp, (perfect body:1.1), realistic, real shadow, 3d, (Swedish forest in the winter background:1.2), (by Michelangelo)
Negative prompt: bad_prompt_version2:0.8
Steps: 25, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 2300539295, Face restoration: CodeFormer, Size: 512x768, Model hash: e04b020012, Model: rpg_V4
```
The prompt templates are good enough that they work well in other models, here's one with DreamShaper.
```
full body, walking pose, slow motion, (rogue ranger:1.2) Swedish female wearing full body multiple stitched layer (fur armor:1.2),(silk cloak), (fabric with intricate pattern:1.2), (insanely detailed, bloom:1.5), (highest quality, Alessandro Casagrande,Greg Rutkowski, Sally Mann, concept art, 4k), (analog:1.2), (high sharpness), (detailed pupils:1.1), (painting:1.1), (digital painting:1.1), detailed face and eyes, Masterpiece, best quality, (highly detailed photo:1.1), 8k, photorealistic, By jeremy mann, by sandra chevrier, by maciej kuciara, sharp, (perfect body:1.1), realistic, real shadow, 3d, (Swedish forest in the winter background:1.2), (by Michelangelo)
Negative prompt: bad_prompt_version2:0.8
Steps: 25, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 1898810504, Face restoration: CodeFormer, Size: 512x768, Model hash: a60cfaa90d, Model: dreamshaper_5BakedVae
```
### Additional resources
* [Artist reference Spreadsheets](https://www.reddit.com/r/midjourney/comments/13k6f75/since_a_lot_of_you_seemed_to_like_my_midjourney/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=share_button)
*
## Bloods and crits
### Dreamspace [on Reddit](https://www.reddit.com/r/StableDiffusion/comments/13hqpgn/dreamspace/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=share_button)
Great repetition of form
Crit: Could go another layer by fixing up the signage. Either make the text, or remove it, or do something iwth the signage that adds to the piece.
### "Remix of my old drawings" [on Reddit](https://www.reddit.com/r/StableDiffusion/comments/13flkcj/really_like_how_the_details_turned_out_in_this/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=share_button)
Look at how well the ideas of the drawing shine through! The floral theme. The interesting composition. The narrative.
There's also this beautiful idea that places with space where it's like the hand is coming out of the frame, which builds space and adds a lot of interest.
from this drawing:
### Streetwear marble statues [on Reddit](https://www.reddit.com/r/StableDiffusion/comments/13i7pg6/streetwear_marble_statues/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=share_button)
I love the sculptural work! And I really like the concept. It's a cool blend of old/new
It does look a little flat, and it has totally centered subjects.
### Mad scientist cat [on Reddit](https://www.reddit.com/r/StableDiffusion/comments/13hi5sd/mad_scientist_cat/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=share_button)
## Control net "reference only"
[Thread on Reddit](https://www.reddit.com/r/StableDiffusion/comments/13gg6g2/new_controlnet_preprocessor_available/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=share_button)
OK, we're going to take a reference image, here's the one I used...
Most of the prompt came from an interrogation of a MJ generation.
Prompt:
```
portrait of a 30yo dutch woman in a walking dress at a lake with a (veil over her head:1.1), 1800s clothing, inspired by Anka Zhuravleva, willow 1 9 8 5, wearing correct era clothes, joongwon charles jeong, chaumet style, looking away from viewer, aesthetic portrait, ponds, miles johnstone, small hat, sorrow, RAW photo, 4k, 8k, film grain, analog style
```
with AnalogMadness v4
Here's what it looks like with the prompt only:
Then, with the `reference_only` (no inpainting, so it's totally raw):
### Another idea with reference_only: use it for skin textures.
[Discussion on reddit](https://www.reddit.com/r/StableDiffusion/comments/13i9w7f/texturing_faces_using_controlnet_reference_only/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=share_button)
### Bonus: Segmentation map control net
It's the one called "seg_"
* See Olivio do it on youtube: https://youtu.be/8wpn7UO3FSk
* [Segmentation map colors on Google Sheets](https://docs.google.com/spreadsheets/d/1se8YEtb2detS7OuPE86fXGyD269pMycAWe2mtKUj2W8/edit?usp=sharing)
There's also a
I was going to use Sargent's painting, [Madame X](https://www.google.com/url?sa=i&url=https%3A%2F%2Fen.wikipedia.org%2Fwiki%2FPortrait_of_Madame_X&psig=AOvVaw3ku3LpySs9m4i6wW1n9ibC&ust=1684515531356000&source=images&cd=vfe&ved=0CBIQjhxqFwoTCPCis8Gr__4CFQAAAAAdAAAAABAd) -- but it didn't love painting.
So I used an existing piece of mine...
And then I ran through the RPG model again with the same prompt as before and I got similar compositions...
## My project update
Really good news for me! One of my pieces is getting a back page on a regional magazine! COOL! I'm pretty psyched.
It's basically a back page image and a couple paragraph blurb about me and the art.
I had an interview this past weekend with the journalist and it stretched me to try explain what the heck is going on when we make.
Also really cool -- It mentions my pack basket LoRA! It doesn't actually say LoRA but that I "teach" the AI programs about the pack basket, which I'm just tickled about.
Sign in with Wallet
Connect another wallet