![pod logo](https://i.imgur.com/SlYH9da.png =600x408) ## Intro Welcome to episode twelve! 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! Little different this week! Today we've got: * Model madness model reviews: 3 models * Bloods and crits: 2 pieces * Technique of the week: * My project update: so you can learn from my process 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. ## Model Madness ### ICBINP Relapse Model * [On Reddit](https://www.reddit.com/r/StableDiffusion/comments/14lro80/icbinp_relapse_now_available/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=share_button) * [On Civitai](https://civitai.com/models/28059) ``` 1920s flapper in a dimly lit nightclub, RAW photo, analog style, 4k, fujifilm xt3 Negative prompt: (bad_prompt_v2:0.8),Asian-Less-Neg,bad-hands-5, BadDream, (skinny:1.2) Steps: 60, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 3506349313, Face restoration: CodeFormer, Size: 576x768, Model hash: b7f578674f, Model: icbinpICantBelieveIts_relapse ``` ![](https://hackmd.io/_uploads/r1aIe_1Y3.jpg) ``` 1990s raver at a warehouse rave, RAW photo, analog style, 4k, fujifilm xt3 Negative prompt: (bad_prompt_v2:0.8),Asian-Less-Neg,bad-hands-5, BadDream, (skinny:1.2) Steps: 60, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 4078248153, Face restoration: CodeFormer, Size: 576x768, Model hash: b7f578674f, Model: icbinpICantBelieveIts_relapse ``` ![](https://hackmd.io/_uploads/B1r6ldJKh.jpg) ### Patiens Ab Arte Model * [On Reddit](https://www.reddit.com/r/StableDiffusion/comments/14mds0r/patiens_ab_arte_my_newest_model/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=share_button) * [On Civitai](https://civitai.com/models/99801/patiens-ab-arte) ``` 1920s flapper in a dimly lit nightclub, RAW photo, analog style, 4k, fujifilm xt3 Negative prompt: (bad_prompt_v2:0.8),Asian-Less-Neg,bad-hands-5, BadDream, (skinny:1.2) Steps: 60, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 1027079759, Face restoration: CodeFormer, Size: 576x768, Model hash: 704c0ccb73, Model: patiensAbArte_v10, VAE: vae-ft-mse-840000-ema-pruned ``` ![](https://hackmd.io/_uploads/HyvVWuytn.jpg) ``` 1990s raver at a warehouse rave, RAW photo, analog style, 4k, fujifilm xt3 Negative prompt: (bad_prompt_v2:0.8),Asian-Less-Neg,bad-hands-5, BadDream, (skinny:1.2) Steps: 60, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 3124615409, Face restoration: CodeFormer, Size: 576x768, Model hash: 704c0ccb73, Model: patiensAbArte_v10, VAE: vae-ft-mse-840000-ema-pruned ``` ![](https://hackmd.io/_uploads/r1d_-u1t3.jpg) ### Absolute Reality Model * [On Reddit](https://www.reddit.com/r/StableDiffusion/comments/14ol8jn/new_version_of_absolute_reality_better_for/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=1) * [On Civitai](https://civitai.com/models/81458?modelVersionId=108576) ``` a 1920s flapper in a dimly lit nightclub, RAW photo, analog style, 4k, fujifilm XT3 Negative prompt: (bad_prompt_v2:0.8),Asian-Less-Neg,bad-hands-5, UnrealisticDream, (skinny:1.2) Steps: 31, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 3776884500, Face restoration: CodeFormer, Size: 576x768, Model hash: e948ca5dc4, Model: absolutereality_v16 ``` ![](https://hackmd.io/_uploads/rJ52bFkK3.jpg) ``` a 1990s euro raver at a warehouse rave, RAW photo, analog style, 4k, fujifilm XT3 Negative prompt: (bad_prompt_v2:0.8),Asian-Less-Neg,bad-hands-5, UnrealisticDream, (skinny:1.2) Steps: 31, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: 2134747874, Face restoration: CodeFormer, Size: 576x768, Model hash: e948ca5dc4, Model: absolutereality_v16 ``` ![](https://hackmd.io/_uploads/SJT8fFkYh.jpg) ## Bloods and crits ### Pirates Ahoy! * [On Reddit](https://www.reddit.com/r/StableDiffusion/comments/14mvig8/pirates_ahoy/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=1) I love this subject matter, so I just love it. And I have to say, it does fit what we'd typically expect for a depiction of a ship of this age. The art would typically showcase a specific ship and show the whole thing within the frame. The render's awesome. The composition does fit what we'd expect for a ship like this. There's post processing done to give it a more analog vibe. It's working. I would say there's more opportunity for narrative here. But overall, it's just rad. I'm not seeing any detail problems, but be on the look out for them on stuff like this that's very involved with say, tons of a ropes and stuff. Could be really neat to generate a bunch in this style and combine them for like customizing a naval battle, a pirate encounter, something like that, could ramp up the drama factor. This is really rad, I'd sub to an insta account of just all rad ships like this tbh. Love pirates. Pirate scenes look REALLY cool when there's a big ship and then it's getting overtaken by small ships, because that's what would've gone down in the golden age of piracy (early 1700s). ![](https://preview.redd.it/3fpgb1tbi49b1.jpg?width=1080&crop=smart&auto=webp&v=enabled&s=2373287dc78a8b24945cbe31f91af72a3806a07a) ### Searching for the past * [On Reddit](https://www.reddit.com/r/StableDiffusion/comments/14oceil/searching_for_the_past/?utm_source=share&utm_medium=android_app&utm_name=androidcss&utm_term=1&utm_content=1) Artist shared their workflow, and used a kind of brew of models. I really like it and it comes out with its own style, which is really nice. I like the mood and the vibe of this as well. I also really like how there's kind of a repetition of form of the womans form, and then the barren trees in the background, that's relaly working. I'm not totally sold on this "box" / "dumpster" as subject matter. It does make me ask what it is, but it's not telling me more about the story. It needs more to integrate it into the scene, there is another like structure in the background which is helping, but I like need more to integrate it into the story and scene. Also could use an upscale and a facial fix, just inpaint and use "inpaint only masked" option to inpaint it as full resolution, should do the trick. Also needs a hand fix. Mood and idea is really good, and has a lot going for it, keep it up. ![](https://i.redd.it/ixcj73g5lg9b1.png) ## Technique of the week * [With Olivio on Youtube](https://www.youtube.com/watch?v=3YusK6wZZuQ&feature=youtu.be) I'm going to use two control nets, because I have a workflow that's kind of a "restore an old school postcard" and turn it into a painting, so I'm going to generally use that... So I started with this line art generation... ![](https://hackmd.io/_uploads/S1J7pOJt2.png) I use that as control net 2... And I wind up with generations like... ![](https://hackmd.io/_uploads/r1e3JtkFh.jpg) From... ``` <lora:add_detail:0.7> a painting of a hotel porch at night, Adirondack chairs, stonework, in the style of meticulously detailed, cottagepunk, packed with hidden details, heavy shading, classical revival, heistcore, manapunk, luggage, vibrant colors in summer time, luscious Negative prompt: (bad_prompt_v2:0.8),Asian-Less-Neg,bad-hands-5, UnrealisticDream, (skinny:1.2) Steps: 31, Sampler: DPM++ 2M Karras, CFG scale: 5, Seed: -1, Face restoration: CodeFormer, Size: 768x576, Model hash: b76cc78ad9, Model: dreamshaper_6BakedVae, ControlNet 0: "preprocessor: lineart_realistic, model: control_v11p_sd15_lineart [43d4be0d], weight: 1, starting/ending: (0, 0.61), resize mode: Crop and Resize, pixel perfect: False, control mode: Balanced, preprocessor params: (512, 64, 64)" ``` Then I add a layer I drew for lighting over the other control net, using it as a reference... ![](https://hackmd.io/_uploads/BJyxeFJF3.png) Then with that applied *NOTE*: I really needed to set the control ending step low, or it was too pronounced. This could be an effect of using two maps. ![](https://hackmd.io/_uploads/rJndxY1F2.jpg) Now let's remove the lineart to see if that helps... ![](https://hackmd.io/_uploads/HyLMZKkYh.jpg) Still kinda meh? It's worth trying here and again, it does give you a little gist of it.