You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The overall goal would be to have ideogram like control over object placement with any model. Here is a current project implementing regional prompting with FLUX.(https://github.com/instantX-research/Regional-Prompting-FLUX) This could be implemented by having prompts be associated with a mask/region and a strength by which their conditioning modifies the current conditioning.
Some of the main challenges I see with this are prompt order would matter in the command line. The text encoder would need to be run multiple times one for each prompt. Region/masking support would need to be added. There would also need to be a strength value to mix in the conditioning of the defined area.(maybe on masks have it be partial strengths; ex 1.0 is full effect, 0.0 is no effect, 0.8 is a 0.8 mix of the new effect). All the different conditioning would need to be be stored(increased vram usage)
What do others think is there another way to accomplish this? If not hopefully this helps provide some ideas on how masking/regions should start to be implemented.
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
Uh oh!
There was an error while loading. Please reload this page.
-
The overall goal would be to have ideogram like control over object placement with any model. Here is a current project implementing regional prompting with FLUX.(https://github.com/instantX-research/Regional-Prompting-FLUX) This could be implemented by having prompts be associated with a mask/region and a strength by which their conditioning modifies the current conditioning.
Some of the main challenges I see with this are prompt order would matter in the command line. The text encoder would need to be run multiple times one for each prompt. Region/masking support would need to be added. There would also need to be a strength value to mix in the conditioning of the defined area.(maybe on masks have it be partial strengths; ex 1.0 is full effect, 0.0 is no effect, 0.8 is a 0.8 mix of the new effect). All the different conditioning would need to be be stored(increased vram usage)
What do others think is there another way to accomplish this? If not hopefully this helps provide some ideas on how masking/regions should start to be implemented.
Beta Was this translation helpful? Give feedback.
All reactions