Changelog

Follow up on the latest improvements and updates.

RSS

We’re introducing SD3 to all paid plans.
What’s New with SD3?
  • Typography: Robust results in typography.
  • Improved performance: Greatly improved performance in multi-subject prompts, image quality, and spelling abilities compared to previous Stable Diffusion models.
  • Smart prompt understanding: Delivers high-quality outputs, even for complex prompts and spatial relationships.
img-4iIezAKNwM72cU5yhu1rb
img-hBp5BzkO90A4BSivhS9oj
We’ve added all three Essential v2 styles to API: Photorealism V2, Artistic V2, and Anime V2, so you can deliver better quality and more detailed images to your apps’ users.
Following Photorealism V2, we updated our Artistic and Anime Essential models!
The following features enhance both Artistic V2 & Anime V2:
  • Enhanced details and textures
  • Improved image resolution (1535x1536 px as a base one)
  • Improved overall image quality
img-yxqUEfq35GSoociFJi3xP
img-xjTMOK0mSV0IQ5ntLjOTc
The image gallery has been improved with a new fullscreen view mode. Click on the image to enlarge it, use arrows to change featured image to the next one.
Zrzut ekranu 2024-05-28 o 16
Image references (Style Reference, Content Reference, and Face Reference (beta)) are now available in all XL models in the Advanced tab.
Photorealism V2 gets new creative addition to the Image to Image 🔥
With Image References, you can select which visual feature from the uploaded image will be injected into the generated output.
Introducing: Style Reference, Content Reference, and Face Reference (beta).
img-o0DcoeDrqiHj6eOUQcsdM
img-yhmpdKgLpE4Csft63srBy
img-0SSc5wxOYy1cAu1MurcuP (1)

new

improved

Photorealism V2

Huge update to the Essential AI Generator's photorealistic style. V2 not only increases native image resolution but also dramatically improves overall quality and prompt understanding.
img-czuBrWDX9Ohv4LxaCOuMH (1)
img-YSolEKqQV7nE7SENvDvYN
img-4pu7vXbB31juKAIUbWNg7
img-bXfr0VEQPfZBI1I56VJ5z

improved

High-Res V2

We have improved the High-Res generation process, resulting in images with much crisper details.
It is available for all Stable Diffusion v1.5-based models (that also includes custom-trained DreamBoth models).
  • Added Image2Image and redesigned the interface for image2image
  • Added Photorealistic Style
  • We have completely refreshed the UI
  • The base model is now SDXL
  • It now generates images with the resolution of 1024x1024 px instead of 512x512
  • You can now download generated images
  • There is a new set of styles now. We removed the old ones and added new ones instead: Base, Anime Illustration, 3D Toon, B&W Sketch
  • We removed the Image2Image function
Load More