## Intro
Welcome to episode thirteen! 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: On FIVE models.
* Bloods and crits: On 3 pieces
* Technique of the week: I've got someone else's to share with you, on creating more unique characters
* My project update:
Bunch of news, a PSA, but no art crits -- I'm late to record, and I was out camping all weekend, and while it was glorious, I am now behind on all my hustles!
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!
### SDXL 0.9 model leaked
It's all over the reddits!
This could be great, but, I can wait it out until 1.0 gets an official drop. It's likely the tooling is going to need changes again too.
### Community model development
Looks like there's someone on Reddit trying to organize some kind of community model / LoRA development!? [This sounds really promising](https://www.reddit.com/r/StableDiffusion/comments/14t8854/revolutionary_idea_lora_training_labs_lets_get/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=1)
I really love the idea of this!
## Model Madness
### Juggernaut
[On civitai](https://civitai.com/models/46422/juggernaut)
Wow. This is a sleeper or something? How'd I not have seen it!?
```
Cover, painting of a 1920s flapper in a night club, wearing a dress, with details, luminism, strip lighting, complex, head and shoulders portrait, 4k concept art portrait by Greg Rutkowski, artgram, WLOP, Alphonse Mucha
Negative prompt: (worst quality, low quality, normal quality, lowres, low details, oversaturated, undersaturated, overexposed, underexposed, grayscale, bw, bad photo, bad photography, bad art)++++, (watermark, signature, text font, username, error, logo, words, letters, digits, autograph, trademark, name)+, (blur, blurry, grainy), morbid, ugly, asymmetrical, mutated malformed, mutilated, poorly lit, bad shadow, draft, cropped, out of frame, cut off, censored, jpeg artifacts, out of focus, glitch, duplicate, (airbrushed, cartoon, anime, semi-realistic, cgi, render, blender, digital art, manga, amateur)++, (3D ,3D Game, 3D Game Scene, 3D Character), (bad hands, bad anatomy, bad body, bad face, bad teeth, bad arms, bad legs, deformities)++
Steps: 30, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 946263007, Size: 616x768, Model hash: 47170319ea, Model: juggernaut_final
```
```
Cover, painting of a 1990s ravers in a night club, wearing a dress, with details, luminism, strip lighting, complex, head and shoulders portrait, 4k concept art portrait by Greg Rutkowski, artgram, WLOP, Alphonse Mucha
Negative prompt: (worst quality, low quality, normal quality, lowres, low details, oversaturated, undersaturated, overexposed, underexposed, grayscale, bw, bad photo, bad photography, bad art)++++, (watermark, signature, text font, username, error, logo, words, letters, digits, autograph, trademark, name)+, (blur, blurry, grainy), morbid, ugly, asymmetrical, mutated malformed, mutilated, poorly lit, bad shadow, draft, cropped, out of frame, cut off, censored, jpeg artifacts, out of focus, glitch, duplicate, (airbrushed, cartoon, anime, semi-realistic, cgi, render, blender, digital art, manga, amateur)++, (3D ,3D Game, 3D Game Scene, 3D Character), (bad hands, bad anatomy, bad body, bad face, bad teeth, bad arms, bad legs, deformities)++
Steps: 30, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 1802497921, Size: 616x768, Model hash: 47170319ea, Model: juggernaut_final
```
## Realistic Vision 4.0
[On civitai](https://civitai.com/models/4201/realistic-vision-v40)
Unsure if it's me, but the natural skin factor is off the charts in this right now.
```
RAW photo, (closeup:1.2), portrait photo of a shy 1920s flapper in a night club, wearing a dress, skin detail, natural skin, dancing, 8k uhd, high quality, film grain, Fujifilm XT3
Negative prompt: (worst quality, low quality, normal quality, lowres, low details, oversaturated, undersaturated, overexposed, underexposed, grayscale, bw, bad photo, bad photography, bad art)++++, (watermark, signature, text font, username, error, logo, words, letters, digits, autograph, trademark, name)+, (blur, blurry, grainy), morbid, ugly, asymmetrical, mutated malformed, mutilated, poorly lit, bad shadow, draft, cropped, out of frame, cut off, censored, jpeg artifacts, out of focus, glitch, duplicate, (airbrushed, cartoon, anime, semi-realistic, cgi, render, blender, digital art, manga, amateur)++, (3D ,3D Game, 3D Game Scene, 3D Character), (bad hands, bad anatomy, bad body, bad face, bad teeth, bad arms, bad legs, deformities)++
Steps: 30, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 2357297661, Size: 616x768, Model hash: afcc6a9cac, Model: realisticVisionV40_v40VAE
```
```
RAW photo, (closeup:1.2), portrait photo of a swedish 1990s raver in a night club, wearing a dress, skin detail, natural skin, dancing, 8k uhd, high quality, film grain, Fujifilm XT3
Negative prompt: (worst quality, low quality, normal quality, lowres, low details, oversaturated, undersaturated, overexposed, underexposed, grayscale, bw, bad photo, bad photography, bad art)++++, (watermark, signature, text font, username, error, logo, words, letters, digits, autograph, trademark, name)+, (blur, blurry, grainy), morbid, ugly, asymmetrical, mutated malformed, mutilated, poorly lit, bad shadow, draft, cropped, out of frame, cut off, censored, jpeg artifacts, out of focus, glitch, duplicate, (airbrushed, cartoon, anime, semi-realistic, cgi, render, blender, digital art, manga, amateur)++, (3D ,3D Game, 3D Game Scene, 3D Character), (bad hands, bad anatomy, bad body, bad face, bad teeth, bad arms, bad legs, deformities)++
Steps: 30, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 3571924704, Size: 616x768, Model hash: afcc6a9cac, Model: realisticVisionV40_v40VAE
```
### Mega model with 4800+ loras baked into it
* [On reddit](https://www.reddit.com/r/StableDiffusion/comments/14vt403/mega_model_with_5400_x_loras_v19_stable_diffusion/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=1)
* [On civitai](https://civitai.com/models/97062?modelVersionId=115930)
I guess we get to ask the question, is more better?
Unsure. It produces nice illustrations, and I think has a strong bend towards anime -- which doesn't surprise me with how much material is used, there's a lot of anime material out there.
I'd give it a shot if you do work on that end of the spectrum.
```
portrait of a 1920s flapper, wearing a dress, skin detail, natural skin, dancing
Negative prompt: (worst quality, low quality, normal quality, lowres, low details, oversaturated, undersaturated, overexposed, underexposed, grayscale, bw, bad photo, bad photography, bad art)++++, (watermark, signature, text font, username, error, logo, words, letters, digits, autograph, trademark, name)+, (blur, blurry, grainy), morbid, ugly, asymmetrical, mutated malformed, mutilated, poorly lit, bad shadow, draft, cropped, out of frame, cut off, censored, jpeg artifacts, out of focus, glitch, duplicate, (airbrushed, cartoon, anime, semi-realistic, cgi, render, blender, digital art, manga, amateur)++, (3D ,3D Game, 3D Game Scene, 3D Character), (bad hands, bad anatomy, bad body, bad face, bad teeth, bad arms, bad legs, deformities)++
Steps: 30, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 3250073338, Size: 616x768, Model hash: 104558a0ca, Model: megaModelWith4800X_20, VAE: vae-ft-mse-840000-ema-pruned
```
Then I added "megan fox" to the words because there's 1.5 million megan fox loras
```
portrait of a 1920s flapper, wearing a dress, skin detail, natural skin, dancing, megan fox
Negative prompt: (worst quality, low quality, normal quality, lowres, low details, oversaturated, undersaturated, overexposed, underexposed, grayscale, bw, bad photo, bad photography, bad art)++++, (watermark, signature, text font, username, error, logo, words, letters, digits, autograph, trademark, name)+, (blur, blurry, grainy), morbid, ugly, asymmetrical, mutated malformed, mutilated, poorly lit, bad shadow, draft, cropped, out of frame, cut off, censored, jpeg artifacts, out of focus, glitch, duplicate, (airbrushed, cartoon, anime, semi-realistic, cgi, render, blender, digital art, manga, amateur)++, (3D ,3D Game, 3D Game Scene, 3D Character), (bad hands, bad anatomy, bad body, bad face, bad teeth, bad arms, bad legs, deformities)++
Steps: 30, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 3270454380, Size: 616x768, Model hash: 104558a0ca, Model: megaModelWith4800X_20, VAE: vae-ft-mse-840000-ema-pruned
```
Then I used a magic prompt...
```
the most gorgeous woman, ernest khalimov body by krista sudmalis, fantasy character portrait, ultra realistic, concept art, intricate details, elegent, digital painting, smooth, sharp focus, illustration, art by artgerm and greg rutkowski and alphonse mucha, artstation
Negative prompt: (bad_prompt_v2:0.8),Asian-Less-Neg,bad-hands-5, BadDream, (skinny:1.2)
Steps: 30, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 4113177340, Size: 616x768, Model hash: 104558a0ca, Model: megaModelWith4800X_20, VAE: vae-ft-mse-840000-ema-pruned
```
### Blank Canvas
* [On reddit](https://www.reddit.com/r/StableDiffusion/comments/14vq9q5/new_super_versatile_model_blank_canvas/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=1)
* [On civitai](https://civitai.com/models/106523/blank-canvas)
I think the name is fitting, it's a pretty fun model, and I might pick it up now and again for some "je ne sais quoi"
```
Portrait of a 1920s flapper in a golden gossamer twinkling gown, up close, 8k, high quality, golden sparkling lighting, hair in a messy bun, beautiful dark fantasy forest background
Negative prompt: (bad_prompt_v2:0.8),Asian-Less-Neg,bad-hands-5, BadDream, (skinny:1.2)
Steps: 30, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 2758772184, Size: 616x768, Model hash: c3fa847723, Model: blankCanvas_v10, VAE: vae-ft-mse-840000-ema-pruned
```
```
Portrait of a 1990s raver, t-shirt and jeans, glowsticks, rave, drum and bass, 8k, high quality, golden sparkling lighting, hair in a messy bun, beautiful rave background
Negative prompt: (bad_prompt_v2:0.8),Asian-Less-Neg,bad-hands-5, BadDream, (skinny:1.2)
Steps: 30, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 2372881298, Size: 616x768, Model hash: c3fa847723, Model: blankCanvas_v10, VAE: vae-ft-mse-840000-ema-pruned
```
### Ordinary Humans
[On civitai](https://civitai.com/models/98755/humans)
This is a cool idea, and I can definitely see a use for it where you need that realism.
Has a "uncanny"-ness to it, that... is probably natural for this kind of a model.
```
Portrait of a 1920s flapper, 8k, high quality, golden sparkling lighting
Negative prompt: (bad_prompt_v2:0.8),Asian-Less-Neg,bad-hands-5, BadDream, (skinny:1.2)
Steps: 30, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 269743286, Size: 616x768, Model hash: 8a3086d0c0, Model: humans_v10, VAE: vae-ft-mse-840000-ema-pruned
```
And bus drivers!
```
Portrait of a 1950s male bus driver, 8k, high quality
Negative prompt: (bad_prompt_v2:0.8),Asian-Less-Neg,bad-hands-5, BadDream, (skinny:1.2)
Steps: 30, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 2622835940, Size: 616x768, Model hash: 8a3086d0c0, Model: humans_v10, VAE: vae-ft-mse-840000-ema-pruned
```
```
Portrait of a 1950s female bus driver, 8k, high quality
Negative prompt: (bad_prompt_v2:0.8),Asian-Less-Neg,bad-hands-5, BadDream, (skinny:1.2)
Steps: 30, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 545463720, Size: 616x768, Model hash: 8a3086d0c0, Model: humans_v10, VAE: vae-ft-mse-840000-ema-pruned
```
## Bloods and crits
### Peek
[On Reddit](https://www.reddit.com/r/StableDiffusion/comments/14p2j84/peek_the_new_patiens_ab_arte_is_a_sleeper_and/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=1)
Using the Patiens Ab Arte from the last episode
Incredible rendering. I mean, crazy good rendering.
Really avoids GPS, the freckles, and the interesting positioning and pose really get all the great qualities of a portrait without getting the generic-factor of it.
The neck seems long, and the body is not great (clothing is uneven, there's this skinniness that looks alien). If it were mine, I'd repaint the bottom portion -- or, honestly, this could use a good crop.
Overall, great work, needs a little iteration to be a final piece, it'd really shine.
### "For a commercial campaign"
[On reddit](https://www.reddit.com/r/StableDiffusion/comments/14tmno6/nearly_unlimited_female_faces/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=1)
Worth loading up to see the artist talk about the process, and what it required from the client.
It's criminal that this isn't upvoted because it's thoughtful and it's clear that this has taken a lot of time to do.
Overall, I like it and I think it's working. I like the HDR-looking colors. The character fits in the scene well. There's nothing major detracting from the piece.
Maybe the koalas are like... overboard? Maybe more focus on like... a few koalas? It's the power of this tool though, you can do A LOT of stuff. Which is awesome. I'd consider taking the right 2/3'd of the image.
I really appreciated what the artist had to do to get the clothing just right for the client. This is very interesting and something to think about as we build our work -- what things must be EXACT, what things can we kind of randomize.
I also like the model -- unsure if real or generated, plays as real, but it doesn't matter, I just like it and it's kind of "non-standard".
### First day at the diner
[On Reddit](https://www.reddit.com/r/StableDiffusion/comments/14x7rzj/first_day_at_the_diner/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=1)
I'm using this one partially because the OP originally said:
> yea and her dress, and fingers. I'm not really big on inpainting etc. Sort of defeats the purpose of AI if I have to spend time fixing stuff.
I had a friend who once said:
> It's a poor artist that blames their tools
So don't. Go ahead and fix it with what you've got. Have a strong vision, and follow through on it. Don't go "oh well, the generation is bad, I'll just leave it that way" -- DON'T.
BUT! Big credit to this person, they changed their mind. And they're going to keep working on it -- huzzah!
It's a really funny two part series. The narrative is all there on this, and it cracks me up. It's provacative, and it's kind of relatable in its own weird way.
It's also weird. But it's weird in way that works.
But it needs touch up to complete it.
It also could use something to kind of "link" these two panels together even further. Yes, the expressions do it -- but if we had some hint from the environment, it would sell it even further.
Think about adding something that's the same in both images. Something to carry the idea.
 
## Technique of the week
[Idea from Reddit](https://www.reddit.com/r/StableDiffusion/comments/14tmno6/nearly_unlimited_female_faces/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=1)
The idea is to use dynamic prompts to help create more unique people ideas.
I'm usually not big on "just making a celebrity" -- that's boring to me because it's a thin-ass concept. I want something different, unique and interesting.
I'll depict a famous person when I need to, and I kind of do for my own project. But... Just doing like "the pope" -- is super boring to me. Or even Megan Fox.
But yeah -- this is a way to kind of develop a set of things that you might find interesting in your characters.
So I decided to
First you're going to need: https://github.com/adieyal/sd-dynamic-prompts.git
I installed it from the extensions list in a1111/vlad.
Alright so I'm creating a few text files... I prompted chatGPT to give me these names.
I asked for names from different time periods.
Now I created a new folder in my wildcards folder...
One file, `f1.txt`
```
Queen Victoria
Susan B. Anthony
Florence Nightingale
Annie Oakley
Sarah Bernhardt
Mary Cassatt
Elizabeth Cady Stanton
Louisa May Alcott
Emily Dickinson
Nellie Bly
Clara Barton
Harriet Tubman
Lillie Langtry
Elizabeth Garrett Anderson
Sojourner Truth
Isabella Bird
Victoria Woodhull
Julia Ward Howe
Mary Todd Lincoln
Madam C.J. Walker
```
Then, `f2.txt`
```
Coco Chanel
Amelia Earhart
Frida Kahlo
Marlene Dietrich
Josephine Baker
Gertrude Stein
Dorothy Parker
Billie Holiday
Helen Keller
Georgia O'Keeffe
Clara Bow
Josephine Cochrane
Louisa May Alcott
Margaret Sanger
Bessie Coleman
Mary Pickford
Annie Oakley
Zora Neale Hurston
Edith Wharton
Eleanor Roosevelt
```
And `f3.txt`
```
Clara Bow
Greta Garbo
Louise Brooks
Joan Crawford
Josephine Baker
Gloria Swanson
Colleen Moore
Marion Davies
Anita Page
Pola Negri
Bessie Love
Theda Bara
Norma Talmadge
Mary Pickford
Olive Thomas
Bebe Daniels
Marion Davies
Clara Kimball Young
May Allison
Phyllis Haver
```
Then a final file, `females.txt` with:
```
[__f1__|__f2__|__f3__]
[__f1__|__f3__|__f2__]
[__f2__|__f1__|__f3__]
[__f2__|__f3__|__f1__]
[__f3__|__f1__|__f2__]
[__f3__|__f2__|__f1__]
```
Then I put together a prompt like this:
```
Cover, painting of __femalefaces/females__, with details, luminism, strip lighting, complex, head and shoulders portrait, 4k concept art portrait by Greg Rutkowski, artgram, WLOP, Alphonse Mucha
```
And ran it with Juggernaut model...
It really gives a lot of character!
And now that you've done that..
## Update on my project.
Having a little bit of a creative block!
And it's coming right after a like... kinda interesting thing that happened where I made a piece, posted it to my social media, and it flopped... Kind of after having a good uptick in how well I'm doing on social media too.
...Even though I really liked the piece personally.
So I have to think about audience.
I make art because I like it and I make what I want to make. But I also want it to be successful with my audience.
I'm trying to work through it, and I might need to do something to get over it (like take a break, or work on another project?). But I'm also going with the "keep on trucking" methodology
Funny I'm feeling like some other stuff I'm making on the side is coming out really well.
And I even kinda tried to use some of that stuff in that failed post....
...so, I gotta look at what does work and follow my patterns, because I'm doing a lot of stuff that's working well for my audience.
Sign in with Wallet
Connect another wallet