Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Opendream: A layer-based UI for Stable Diffusion

github.com

Post reply on HN
Print view View original

1–10 of 158 posts

Re: Opendream: A layer-based UI for Stable Diffusion

#2

Post by cwkoss »

Very cool. Would be interesting to train a model on images with alpha channels so outputs would be automatically masked and more easily composable. But maybe masking is so good these days that would be futile?

When a user does img-2-img on a layer does it use the context from other visible layers in the generation?

Re: Opendream: A layer-based UI for Stable Diffusion

#3

Post by dheera »

cwkoss wrote: post #2

Very cool. Would be interesting to train a model on images with alpha channels so outputs would be automatically masked and more easily composable. But maybe masking is so good these days that would be futile? When a user does img-2-img on a layer does it use the context from other visible layers in the generation?

For composing this approach works pretty well, maybe the author should consider making a UI for it

https://multidiffusion.github.io/

Re: Opendream: A layer-based UI for Stable Diffusion

#4

Post by Zetobal »

cwkoss wrote: post #2

Very cool. Would be interesting to train a model on images with alpha channels so outputs would be automatically masked and more easily composable. But maybe masking is so good these days that would be futile? When a user does img-2-img on a layer does it use the context from other visible layers in the generation?

Segmentation is solved... https://github.com/RockeyCoss/Prompt-Segment-Anything

Re: Opendream: A layer-based UI for Stable Diffusion

#5

Post by mdp2021 »

cwkoss wrote: post #2

Very cool. Would be interesting to train a model on images with alpha channels so outputs would be automatically masked and more easily composable. But maybe masking is so good these days that would be futile? When a user does img-2-img on a layer does it use the context from other visible layers in the generation?

> Would be interesting to train a model on images with alpha channels

Would be even more interesting to get an ANN middle system of ontology of the (finally) represented content in order to change the single items.

An internal representation of qualified structured items in space as part of the chain. Prompt > accessible internal representation > render.

Post reply on HN