# This is not an AI Art Podcast (Ep. 7)
## Intro
Welcome to episode seven! 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:
* "Let's play art teacher": Where I take pieces of AI and work on them without your permission lol!
* Technique of the week:
* 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
[New nvidia driver out](https://www.reddit.com/r/StableDiffusion/comments/13q4ku4/nvidia_2x_performance_improvement_for_stable/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=share_button) -- claiming a double speed for SD, but, we havent' seen it yet, let's track it.
[Fondant dataset gathering OSS project](https://github.com/ml6team/fondant)
Seems like a cool idea!
## Model Madness
### DreamShaper v6
This model is great and I'm often picking it up, especially to touch up other painterly and illustrative works.
Also pick up the [BadDream and UnrealisticDream](https://civitai.com/models/72437/baddream-unrealisticdream-negative-embeddings) negative embeddings.
```
painting of a 1920s flapper in a night club, dramatic lighting, dancing, party clubbers, detailed faces, 1920s, donalds, David Hockney, by Viktor Vasnetsov, concept art, fantasy, vibrant, hd shot, digital portrait, beautiful, artstation, comic style, by Artgerm, guy denning, jakub rozalski, magali villeneuve and charlie bowater
Negative prompt: BadDream
Steps: 25, Sampler: DPM++ 2M Karras, CFG scale: 7, Seed: -1, Face restoration: CodeFormer, Size: 768x576, Model hash: b76cc78ad9, Model: dreamshaper_6BakedVae
```
```
photograph of a 1920s flapper in a night club, dramatic lighting, dancing, party, RAW photo, 4k, 8k, UHD, Fujifilm 300 XT trending on ArtStation
Negative prompt: BadDream, (UnrealisticDream:1.2)
Steps: 25, Sampler: DPM++ 2M Karras, CFG scale: 7, Seed: -1, Face restoration: CodeFormer, Size: 768x768, Model hash: b76cc78ad9, Model: dreamshaper_6BakedVae
```
```
a swedish raver woman at a rave, dancing, 3 0 years old, hot summer night, 90s clothes, bandana, candy necklace, chunky jewelry, glowsticks
Negative prompt: BadDream, UnrealisticDream, asian
Steps: 25, Sampler: DPM++ 2M Karras, CFG scale: 7, Seed: 4261327939, Face restoration: CodeFormer, Size: 576x768, Model hash: b76cc78ad9, Model: dreamshaper_6BakedVae
```
### CyberRealistic
On civitai: https://civitai.com/models/15003/cyberrealistic
```
color photograph, ((a realistic photo of a1920s flapper)), 1920s dress, (milalc), light, ((glowy skin)), looking_at_viewer, (fit body:1.0), detailed illustration, masterpiece, high quality, realistic, very detailed face,
Negative prompt: bad_prompt_version2:0.8, bad-hands-5, asian
Steps: 25, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 2071125106, Face restoration: CodeFormer, Size: 512x768, Model hash: 661697d235, Model: cyberrealistic_v30
```
### I can't believe it's not photography
https://civitai.com/models/28059?modelVersionId=76459
```
color photograph of a 1920s flapper dancing at a nightclub, 1920s dress, bar, lights, bokeh, masterpiece, best quality, ultra-detailed, (skin texture) (film grain:1.3), (warm hue, warm tone) :1.2), close up, cinematic light, sidelighting, ultra high res, best shadow, RAW
Negative prompt: bad_prompt_version2:0.8, bad-hands-5, asian, monochrome, black and white
Steps: 25, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 878376493, Face restoration: CodeFormer, Size: 512x768, Model hash: c26f4c4227, Model: icbinpICantBelieveIts_v7
```
### Spaceship LoRA
It's fun! I'm not getting much for ability to vary it, but I like the generations.
```
<lora:sp4c3sh1p:0.8> 1950's sci-fi spaceship [sp4c3sh1p:0.3], outerspace, moon craters <lora:add_detail:1>
Negative prompt: bad_prompt_version2:0.8, bad-hands-5, BadDream
Steps: 25, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 3520956820, Face restoration: CodeFormer, Size: 768x512, Model hash: b76cc78ad9, Model: dreamshaper_6BakedVae
```
### Age Slider embedding
https://civitai.com/models/65214/age-slider
### Marble statue LoRA
https://civitai.com/models/70538/marble-make-everything-into-marble
### 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)
## Additional Resources
* [LAION search demo](https://rom1504.github.io/clip-retrieval/?back=https%3A%2F%2Fknn.laion.ai&index=laion5B-H-14&useMclip=false&query=Bourguereau+)
Check out what your prompts might be pulling from the dataset.
* [ComfyUI Fedora installation](https://www.reddit.com/r/StableDiffusion/comments/13n9prv/linux_fedora_38_comfyui_installation_in_5_minutes/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=share_button)
I'm hopeful to share some of my Linux recipes as well!
* Some nice photo prompt templates
* [From Reddit](https://www.reddit.com/r/StableDiffusion/comments/13ku6y5/trying_to_create_a_feeling_of_intimacy/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=share_button)
*
## Bloods and crits
### Exquisite Surrealist Floral Princess
Not sure I agree with the "surrealist" label on this, but floral princess yes.
Cool generation and thanks for sharing the prompt! A lot of detail went into the prompt. Generation & rendering: great. Composition and achieving narrative: not so great.
From [this reddit post](https://www.reddit.com/r/StableDiffusion/comments/13qy7kt/exquisite_surrealist_floral_princess/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=share_button)
We'll play art teacher later.
### Forest [on Reddit](https://www.reddit.com/r/StableDiffusion/comments/13omhod/forest/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=share_button)
Asymmetrical balance on this really drew me to it.
It's really good overall, but it could probably use an upscale. There's nothing in particular that's detracting from it.
### The date [on Reddit](https://www.reddit.com/r/StableDiffusion/comments/13o1hae/the_date/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=share_button)
This looks great overall. It came out really well. It's so close, but probably it'd be worth doing a pass where you fix the things that are out of place.
* The signage (make it intentional or remove it)
* There's a wrist watch that's messed up
* There's necklaces that don't make sense
* Also the left eye is a bit off
* Strong yello vertical line on the right is detracting from the piece
## Let's play art teacher
Where I take your art and manipulate it.
From [this reddit post](https://www.reddit.com/r/StableDiffusion/comments/13qy7kt/exquisite_surrealist_floral_princess/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=share_button)
We start with this, which is a beautiful generation -- but it's got a few problems...
* The composition boring. It has a kind of GPS composition, but not necessarily this like, GPS where there's no narrative. I think we have a good start on the narrative
* The prompt is for a cyborg with brown hair. Is this a cyborg?
So, this didn't really read as "cyborg" to me. And the composition needs work. Let's see what we can do...
First I borrowed the prompt. Thank you for posting it!
You can also see that I've decided to change the aspect ratio.
I tested it and got:
```
beautiful cyborg with brown hair, intricate, elegant, highly detailed, majestic, digital photography, (art by artgerm and ruan jia and greg rutkowski, flowers of hope by Jean-Honor Fragonard, Peter mohrbacher), surreal painting gold filigree, broken glass, ornate frame:0.3, jewelry, hyper detailed, insane details, stunning, intricate, elite, art nouveau, ornate, liquid wax, elegant, luxury, Greg Rutkowski, ink style, sticker, vector-art beautiful character design, double exposure shot, luminous design, flowers in hair:0.45, head piece:0.15, circlet, seductive, perfect body, realistic metals, (masterpiece, award winning, sidelighting, finely detailed beautiful eyes: 1.2), hdr, (off angle), ((analog style)), (photorealistic:1.4), (skeleton like:0.42), simple mottled background, long auburn red hair, (concept art)
Negative prompt: bad_prompt_version2:0.8, bad-hands-5, BadDream
Steps: 25, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 1296095376, Face restoration: CodeFormer, Size: 1024x768, Model hash: b76cc78ad9, Model: dreamshaper_6BakedVae
```
I decided I need to trim that prompt down so I have more control over it, so I remove... a good chunk of it and see that I'm still getting similar style from the artist references.
I think it works.
```
beautiful cyborg with brown hair, intricate, elegant, highly detailed, majestic, digital photography, (art by artgerm and ruan jia and greg rutkowski, flowers of hope by Jean-Honor Fragonard, Peter mohrbacher),
Negative prompt: bad_prompt_version2:0.8, bad-hands-5, BadDream
Steps: 25, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 1998138063, Face restoration: CodeFormer, Size: 1024x768, Model hash: b76cc78ad9, Model: dreamshaper_6BakedVae
```
I'm still getting these kinda golden clad flower girls.
Now, let's photobash this thing.
I want to play on this idea of "cyborg" + "flower girl" so I go and collect some borrowed imagery from google images.
I then laid out a circuit board (that I matched the pallette to the original) and then took white and yellow flowers added them to the circuit board. And put some circuit board over here, because I want to push the cyborg narrative.
I had to change the prompt because the prompt is SO opinionated, there was no way we're getting a cyborg at all out of this thing.
```
the woman with a circuit board face is connected to the flower patch, eye visor, (circuitboard face:1.2), intricate, elegant, highly detailed, majestic,
Negative prompt: bad_prompt_version2:0.8, bad-hands-5, BadDream, asian
Steps: 25, Sampler: Euler a, CFG scale: 6, Seed: 238494489, Size: 1024x768, Model hash: b76cc78ad9, Model: dreamshaper_6BakedVae, Denoising strength: 0.57, Mask blur: 4
```
NOTE: I inpainted a few times and adjusted denoising to varying amounted.
I wound up with:
## My project
### Training notes
Did a large LoRA run at 100k steps. Settings outside of default
Number of images: 500
Steps per image: 100
Network Rank (dimension): 128
Batch size: 6
Epochs: 16
Max resolution: 768,768
Base model: Analog Madness v1
It came out "ok" -- I get the effect I want for the most part. But I have to use it at a low weight, which might be normal for a LoRA with that many steps.
I started another one (mostly same but 800 images in training set) but this time I got staticy images, even with a weight of `0.01` the only thing I can see that I changed was that I changed:
Network alpha: 128
Because I saw a "Life is boring, so programming" YT video that said you could make these two match.
I did see this error though:
```
fatal: detected dubious ownership in repository at '/app'
To add an exception for this directory, call:
git config --global --add safe.directory /app
```
### Photshop A1111 plugin
Tutorial from Olivio: https://youtu.be/Y3KJli8ohKI
Sign in with Wallet
Connect another wallet