Regrettably, we're retiring three of our original algorithms: Style Transfer, Artistic (VQGAN+CLIP), and Coherent (CLIP-Guided Diffusion). These models will be removed from NightCafe on December 1, 2026.
Our GPU compute provider is sunsetting the platform these legacy models run on. This is outside our control, and we're truly sorry to see them go.
If you used these models for text-to-image, most current models can recreate some of the effect with keywords like "dreamlike, surreal, incoherent, swirly, disco diffusion render" etc.
If you used Style Transfer to restyle photos, newer models are far better at this than Style Transfer ever was. Use the new Image Editing Model Comparison tool and take a look at the "Style Transfer" category.
Style Transfer is where NightCafe began. In 2020, turning a photo into a painting was the whole product, a small experiment that grew into a community. Artistic (VQGAN+CLIP) was our first text-to-image algorithm (and NightCafe was the first text-to-image website at the time). For the first time, you could type a sentence and watch something impossible appear. Coherent followed, giving that early magic more structure and control.
These models are not just old software. They are the roots of NightCafe. Countless first creations, late-night experiments, and pieces that still hang on bedroom walls started here. We know some of you still come back to them for that particular look. The slightly wild, unmistakably early-AI feel that newer models have moved on from.
It is genuinely sad to close this chapter. We cannot keep running them once our GPU compute provider sunsets the platform they live on. Your existing creations stay. The community those models started is still here. And the newer models in Studio exist because these ones showed us what was possible.
Thank you for making art with them. We'll miss them too.