Stream-of-Consciousness-SDXL-v1.0.safetensors

218 MB · SafeTensor

Requires membership

<stream-conscious-xl>

Stream-of-Consciousness Laconic: Stream-of-Consciousness works to create unusual and surprising compositions for your generation results to help create absurd, surreal and interesting collages, illustrations, portraits and so on. Loquacious: Stream-of-Consciousness is a LoRA trained off of Jason Paschal's hand-made absurdist-surrealist collage series of the same name. As a concept it involved gathering a wide variety of images and bringing them together without - or at times deliberately against - having a cohesive message to convey. It's only after a piece was completed that the attempt to find meaning began. Providing the most tangential connections, knowingly employing logical fallacies and making wildly spurious claims, the rhetoric would become reality... at least in that people just accepted that all that artistic posturing is usually all for an idiotic punchline. While the idea was to at the very least emulate the kind of collages that exist in the series the hope was to also bring the spirit of the series to all kinds of other compositions and images. Art as entertainment and art as a joke without diminishing everything else art is. While all of those takes on art are subjective, I'm going to be all self-congratulatory and say that I feel like it does exactly what I was hoping for. This is a total and complete victory - for all of us. Someone pour champagne on the babies. The way I format my prompts when using this LoRA is pretty formulaic and has worked really well thus far: * embeds, trigger * a handful of concrete visual nouns * a primary subject (optional) * style, medium Not that different from any other prompting setup, as you see. However I feel like having the relationship between the objects and the subject and other objects being ambiguous gives cooler results. I use wildcards to randomize the objects themselves but choosing items thematically or adding whatever you see when you look around your room are both valid choices. If you're stuck for an idea, here's a short list to kickstart things: whale, bone, tentacle, skyscraper, picture frame, waterfall, mushroom cloud, phonograph, seashell, cherry blossom petal, floppy disk, projector screen, glass shard, kestrel wing, sewing machine, seahorse tail, giant death laser, zebra mask, zodiac symbol, steam vents, swarm of bees, CRT TV, yellow pencil, ashtray, gravitational wave, satellite dish, flip phone, slot machine, snakes, punching bag, violin bow, storm cloud, lower intestines, astronaut helmet, animal skull, floating trail of blood, broken electronics, chessboard, ethernet cable, ink bottle, stethoscope, starry night sky, razor, wading boots, conch shell, swivel chair, flaming skull, dentures, mechanical spine, hydraulic pistons, magic wand, milk, zoetrope Style(s) then medium(s) in and unbroken prompt together has worked best for me, though they're technically entirely optional. If you are using them, adding another set styles/media afterwards if you want and then adding other descriptors can bring more cool, chaotic results. Styles and mediums are a bit more hit-and-miss in Illustrious I've come to discover but they're not absent entirely (making versions for other models is probably on my short list). This LoRA is suited to, if nothing else, experimentation. All that said, here's a few medium/style prompts that came back with interesting results for me: * punk rock torn paper collage \(medium\) with misprinted cmyk screentones + scotch tape, distressed textures, blacklight poster, inverted colors, * isometric holographic sculpture with laser-cut paper \(medium\) + mylar reflections, shadowbox, datamoshing, livestreaming, * absurdist 4 koma \(manga\), pop surrealism collage \(style\), hand-cut paper collage, ink drawing, halftone, greyscale, monochrome, manga panel, desaturated, * professionally printed concept art technical documents, feminist illustration, anime \(style\), retro video game box art \(medium\), * stoner art, pop art black light painting on black velvet \(medium\), inverted colors, neon glow, shiny iridescent holographic poster art, I'd love to hear any combos other people stumble across or any other discovery in how to use the LoRA from anyone that wants to share. Make something subjectively beautiful and unleash it onto the unsuspecting world. And next time someone decides to say AI can't make art because it "takes no effort" ask them where the hell it says that making art is supposed to be hard.

SHA256

99D52C56C8751CBFA3F48F7A53021D87F75A28B8E263EC30926D06C353175EC7

ModelTrends.ai Model ID

#20296

Sampler

Euler a

Steps

20

Guidance

7

Seed

1275044344

Size

1024x1536

Sampler

Euler a

Steps

20

Guidance

7

Seed

3231896420

Size

1024x1536

Sampler

DPM++ 2M SDE

Steps

16

Guidance

3.5

Seed

1975045973

Size

1024x1536

Sampler

DPM++ 2M

Steps

30

Guidance

3.5

Seed

331601049

Size

1024x1536

Sampler

Euler a

Steps

20

Guidance

7

Seed

532671603

Size

1024x1536

Sampler

Euler a

Steps

20

Guidance

7

Seed

818949882

Size

1024x1536

Sampler

DPM++ 2M SDE

Steps

16

Guidance

3.5

Seed

3329159507

Size

1024x1536

Sampler

Euler a

Steps

16

Guidance

3.5

Seed

3849563655

Size

1024x1536