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
I'm having an issue with saved image's workflows having the wrong saved prompt. When running a batch count job or just queueing jobs, the saved workflow prompt for all images is the very first prompt generated instead of whatever the current prompt happens to be. Also for singular generation, the generated prompt may be off by one in a forward manner like the very first generated image will have no saved prompt, the next generated image will have the prompt of the previous image, and so on. There is also a minor issue of when changing the tags, the node needs to be reloaded which will clear the tag_filter the first time the node is reloaded but not on any subsequent reloads. Other than not being able to recreate a randomly generated prompt because the saved prompt information in the workflow is often inaccurate, this still works exceptionally well for testing checkpoints and LORAs. Here is the workflow I am using below, but this also occurred when using the workflow provided on the Github and swapping out Derpibooru with Danbooru. I don't actually know if this occurs when using the Derpibooru or E621 generators instead of Danbooru though as I'm weeb.
The text was updated successfully, but these errors were encountered:
I'm having an issue with saved image's workflows having the wrong saved prompt. When running a batch count job or just queueing jobs, the saved workflow prompt for all images is the very first prompt generated instead of whatever the current prompt happens to be. Also for singular generation, the generated prompt may be off by one in a forward manner like the very first generated image will have no saved prompt, the next generated image will have the prompt of the previous image, and so on. There is also a minor issue of when changing the tags, the node needs to be reloaded which will clear the tag_filter the first time the node is reloaded but not on any subsequent reloads. Other than not being able to recreate a randomly generated prompt because the saved prompt information in the workflow is often inaccurate, this still works exceptionally well for testing checkpoints and LORAs. Here is the workflow I am using below, but this also occurred when using the workflow provided on the Github and swapping out Derpibooru with Danbooru. I don't actually know if this occurs when using the Derpibooru or E621 generators instead of Danbooru though as I'm weeb.
The text was updated successfully, but these errors were encountered: