Luma Labs and Kling may be generating buzz, but StabilityAI has just released its latest text-to-image model, SD3, which can be run locally at home—and it just came out yesterday! Features & Misses ✅Improved photorealism (Images 1-4) ✅Improved prompt adherence and type support (Image 5-6) ❌Human and animal anatomy can go all sorts of wrong ❌SD3 license is restrictive compared to SDXL and 1.5’s CreativeML License ❌The new model is receiving lots of negative feedback from the community Render time ~5.5s to generate a 1024x1024 image using the non-tensor model Models and Installation 📦SD3 Models: https://lnkd.in/et6nn29x ⚙️ComfyUI: https://lnkd.in/efwbUmBF Image 5 prompt: 2 jars, the first one is large and has a jellyfish and rocks in it with the word "Jellyfish" etched into the glass, and the other jar is small and has gold dust in it with the word "Gold" etched into the glass, minimalist background, teal blue background, plant in the background
-
-
-
-
-
+1
