Replies: 2 comments
-
In live mode only the area covered by the region layer is changed. The rest of the image stays the same. This is indicated by the grid overlay. Live mode doesn't render the whole image with regions like the Generation workspace does, it's too slow. See https://docs.interstice.cloud/regions/#live-mode Regarding groups, I tried to explain here: https://docs.interstice.cloud/regions/#layer-groups |
Beta Was this translation helpful? Give feedback.
-
Thank you for clarifying. Amazing tool. |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Thank you for the plugin.
Disclaimer: this is not critisism of any kind to your outstanding work, I am very grateful for the artistic freedom you provide. I understand that you are not a Comfy dev nor a Krita dev, that products kan work different than documented.
There is a "problem" with local prompts and live mode. When busy on a local prompt then only a part of the screen is shown as editable, the rest of the preview screen gets "hidden" behind an ugly grid, and in the live update it's not the image+layers that gets rendered but the selection plus what appears to be that malformed "gridded" full picture. For example, I can start with a landscape background, it renders well. Then I add a layer for a house: the main image becomes a grid, the preview becomes a mess instead of the original image plus the house getting drawn in the front. It's just, I have a background, and invoking regional prompt without doing anything already changes the preview. Maybe I am missing something.
I understand SD takes everything in account, but an empty layer should not change the output, (nor should control layer images imo, do not send them as control layers + img2img layers, or give an option).
Also, collapsing into an image can result in a destructive layers = 1 group mess where 'undo' does nothing but go step by step back many steps. It's easy to get lost after a stupid-user (=me) move. Layers are merged and grouped somehow. I know there are extra optons now but I don't understand how it works. Even with Grok as assistant and saving many times it's difficult. I simply don't know what is happening and why, making art with basic functionality just works as expected (and again thank you).
This is not a complaint or a list of demands or something, your plugin is awesome. It's just not always intuitive. Via trial and error I found a pretty optional environment with several 1.5 models and specific settings to quickly get what I want while using Line Art and sometimes Scribble (using the original drawing, not some generated controlnet picture that blends 100% in), but regional prompts never worked for me. Maybe regional prompts could be layers too, without extensive tweaking. For example. layer 1 = background mounains, layer 2 = woods, layer 3 = house. No grids blocking the canvas from view and rendering the end result, no control images that also function as img2img.
Again, thank you very much. Idf like to support but only IDeal is possible here in Holland (No credit card), please share if there is another support option. The plugin as is is worth $20 at least IMO.
Thanks.
Beta Was this translation helpful? Give feedback.
All reactions