Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
CamperBob2
6 hours ago
|
parent
|
context
|
favorite
| on:
Qwen-Image-Layered: transparency and layer aware o...
That's what the example code at
https://old.reddit.com/r/StableDiffusion/comments/1pqnghp/qw...
generates. You get 0.png, 1.png ... n.png, where n= the requested number of layers-1.
It'll drop a 600W RTX 6000 to its knees for about a minute, but it does work.
dvrp
5 hours ago
[–]
I saw some people at a company called Pruna AI got it down to 8 seconds with Cloudflare/Replicate, but I don't know if it was on consumer hardware or an A100/H100/H200, and I don't know if the inference optimization is open-source yet.
reply
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search:
It'll drop a 600W RTX 6000 to its knees for about a minute, but it does work.