from Hacker News

Generating transparent images using Stable Diffusion XL

by tracyhenry on 3/3/24, 12:27 AM with 13 comments

• by amluto on 3/3/24, 1:30 AM

  Looking at the “woman, messy hair, high quality” image, the hair farther from her head looks wrong in much the way that iPhone portrait mode messes up hair. I wonder if this is an example of an AI training on partially AI-generated data and reproducing its artifacts.

• by GaggiX on 3/3/24, 1:09 AM

  Paper: https://arxiv.org/abs/2402.17113

  The author Lvmin Zhang is the same person behind ControlNet.

• by vunderba on 3/3/24, 1:52 AM

  The partial alpha blending support for translucent materials is really cool (glass, plastic, etc).

  I'd be curious to see how well this plays with inpainting. Apparently img2img is also on the authors todo list.

• by swyx on 3/3/24, 12:57 AM

  reactions

  1 - the way the dog at the end gets a reflection off the floor is pretty nice.

  2 - i wonder how this compares in terms of latency/complexity with a comfyui pipeline that just does a typical edge detection/masking layer to achieve the transparency effect. however i dont think that method would work with the glass example as shown

• by dannyw on 3/3/24, 12:42 AM

  Apache 2.0, the beauty of open source. Nice.