AI model Stable Diffusion upgrade: normal generation of five fingers, more realistic image

Thanks to CTOnews.com netizen OC_Formula for the clue delivery! CTOnews.com June 24 news, Stability AI recently issued a press release, announced the release of SDXL version 0.9 update, upgraded the Stable Diffusion text generation picture model.

In the press release, Stability AI also compares the version of SDXL Beta released in April this year with the update of version 0.9. The comparison figure in the attached blog post of CTOnews.com is as follows, with SDXL Beta on the left and SDXL 0.9 on the right:

It is obvious that the pictures generated by the new version are more realistic and textured, and can correctly generate five human fingers. Stability AI said that the configuration requirements for SDXL version 0.9 are the same as before, requiring 16GB memory and Nvidia RTX 20 series and higher video cards (8GB and above video memory), and supporting Win10, Win11 and Linux distributions.

Stability AI says users can access the model through ClipDrop, and the associated API is about to be released, which is expected to be released in mid-July.