Flux 1.0Longwinded explanation will follow, maybe, 'cause Flux is fluxing.<hr />A little helper for weird looking pictures that are a bit more on the interesting and don't reveal everything at the first glance.<ul><li>Adds weird geometry and optical illusions</li><li>Changing just one word, no matter if it makes sense or not, can change everything</li><li>Even nonsensical prompt combinations could lead to something</li><li>A weight like 0.8 or lower is recommended because there are a lot of different styles combined in the dataset.</li></ul>There was no intention behind this LoRA apart from me getting a bit bored and testing a few things. Like with all my LoRAs i will sooner or later find a good use for it. Combing even the most ridiculous models that shouldn't fit together is the best thing.

Flux.1 D / SDXL - Impossible Geometry

- Flux1 - impossible_geometry_V1.0.safetensors

Flux.1 D - v1.0

SDXL v1.0

Longwinded explanation will follow, maybe, 'cause Flux is fluxing.A little helper for weird looking pictures that are a bit more on the interesting and don't reveal everything at the first glance.<ul><li>Adds weird geometry and optical illusions</li><li>Changing just one word, no matter if it makes sense or not, can change everything</li><li>Even nonsensical prompt combinations could lead to something</li><li>A weight like 0.8 or lower is recommended because there are a lot of different styles combined in the dataset.</li></ul>There was no intention behind this LoRA apart from me getting a bit bored and testing a few things. Like with all my LoRAs i will sooner or later find a good use for it. Combing even the most ridiculous models that shouldn't fit together is the best thing.

Impossible Geometry Flux.1 Dev

Longwinded explanation will follow, maybe, 'cause Flux is fluxing.



A little helper for weird looking pictures that are a bit more on the interesting and don't reveal everything at the first glance.







Adds weird geometry and optical illusions



Changing just one word, no matter if it makes sense or not, can change everything



Even nonsensical prompt combinations could lead to something



A weight like 0.8 or lower is recommended because there are a lot of different styles combined in the dataset.



There was no intention behind this LoRA apart from me getting a bit bored and testing a few things. Like with all my LoRAs, I will sooner or later find a good use for it. Combing even the most ridiculous models that shouldn't fit together is the best thing.

Files with SHA256: dc33547ffc2809c5f54f8085cd2c5cec97756165eb69c5981ec4d3614833a2fc