← sources

Visual Anagrams with Iris 3B Turbo on an RTX 3060 (12 GB)

source: reddit (reddit:StableDiffusion)published 2026-10-24 03:34ingested 2026-10-24 10:59

retained text

The original Visual Anagrams used DeepFloyd IF, a pixel-space diffusion model.

I adapted it to use Iris 3B with INT8 weights and the community Turbo r2 LoRA. These examples were generated locally at 1024×1024 in four denoising steps on an RTX 3060 with 12 GB VRAM.

The video shows different transformations and styles, including rotations, flipping, color inversion, and patch rearrangement. Results are experimental, and some prompt pairs work better than others.

[https://github.com/PasiKoodaa/iris\_visual\_anagrams](https://github.com/PasiKoodaa/iris_visual_anagrams)

document fetched 2026-10-25 00:30 · canonical https://reddit.com/r/StableDiffusion/comments/1xdy81f/visual_anagrams_with_iris_3b_turbo_on_an_rtx_3060