UI enhancement: When recalling a promt, open the corresponding image … #403
Conversation
…in the image editor when there is an image with the same name as the seed value in the promt.
operators/view_history.py
Outdated
| image_names_no_suffix = [i.name.split('.')[0] for i in bpy.data.images] | ||
| image_names_no_suffix_seed_count = image_names_no_suffix.count(seed_string) | ||
| if image_names_no_suffix_seed_count == 1: |
There was a problem hiding this comment.
Does an exact match catch it correctly:
| image_names_no_suffix = [i.name.split('.')[0] for i in bpy.data.images] | |
| image_names_no_suffix_seed_count = image_names_no_suffix.count(seed_string) | |
| if image_names_no_suffix_seed_count == 1: | |
| if (existing_image := bpy.data.images.get(seed_string, None)) is not None: |
There was a problem hiding this comment.
Unfortunately the proposed change is not equivalent. Consider the following situation. We have made three runs with constant seed of 1096871657 because we wanted to see how the result looked with different samplers. Thus we have three images: 1096871657, 1096871657.001 and 1096871657.002.
Now I want to recall the second promt. With your change, it would select the first of the three images, which is the wrong one. Unfortunately we do not know which of the three is the right image with just the seed. That's why I have decided to count the occurences of a seed in the prefixes of the names of all image datablocks (line 46 and 47) and only select an image if it is the only one with that seed as prefix (line 48).
When there is more than one occurence, it does not change the current image but reports that more than one has been found (lines 54 and 55).
I am aware that this is a heuristic and not perfect but I think I have an idea for a better solution. The problem is that the unique identifier for datablocks in Blender is the name, which is prone to be changed by the user and sideffects of Blender (.001 .002 .003 etc.). What we can do is the following: After an image is created, run hash() over the datablock and store the result as a custom property of said datablock and store it as well in dream_textures_prompt . That way we can definitely select the correct image no matter how it is named.
operators/view_history.py
Outdated
| for area in context.screen.areas: | ||
| if area.type != 'IMAGE_EDITOR': | ||
| continue | ||
| area.spaces.active.image = bpy.data.images[seed_string] |
There was a problem hiding this comment.
This goes with the suggestion above if that works for you.
| area.spaces.active.image = bpy.data.images[seed_string] | |
| area.spaces.active.image = existing_image | |
| self.report({'INFO'}, "Existing generated image opened in image editor") |
operators/view_history.py
Outdated
| elif image_names_no_suffix_seed_count > 1: | ||
| self.report({'INFO'}, "More than one image found for seed " + seed_string + ". Not changing image editor contents.") | ||
| elif image_names_no_suffix_seed_count == 0: | ||
| self.report({'INFO'}, "No image found for seed " + seed_string + ". Not changing image editor contents.") |
There was a problem hiding this comment.
I think I would rather show no message if an existing image isn't found, because users may think this is an error. I think it may be common to try reloading a prompt from a file that isn't open, so it may error more often than it doesn't.
| elif image_names_no_suffix_seed_count > 1: | |
| self.report({'INFO'}, "More than one image found for seed " + seed_string + ". Not changing image editor contents.") | |
| elif image_names_no_suffix_seed_count == 0: | |
| self.report({'INFO'}, "No image found for seed " + seed_string + ". Not changing image editor contents.") |
operators/view_history.py
Outdated
| image_names_no_suffix = [i.name.split('.')[0] for i in bpy.data.images] | ||
| image_names_no_suffix_seed_count = image_names_no_suffix.count(seed_string) | ||
| if image_names_no_suffix_seed_count == 1: | ||
| print("Image for seed " + seed_string + " is unique. Selecting image datablock.") |
There was a problem hiding this comment.
Possibly remove the print statement to keep the logs clean.
| print("Image for seed " + seed_string + " is unique. Selecting image datablock.") |
…on can later be used to open the right image datablock in the Blender image editor when recalling the promt
… in the promt history as well. When recalling a prompt, load the image in the image editor if available
|
I changed the way it works completely. Now a hash is created for each image and stored in a custom property of said image and in the promt history. When recalling a prompt from history and an image with said hash can be found in byp.data.images, show it in the image editor. |
Co-authored-by: Carson Katri <Carson.katri@gmail.com>
carson-katri#403) * UI enhancement: When recalling a promt, open the corresponding image in the image editor when there is an image with the same name as the seed value in the promt. * Add a hash of the resulting image as string to prompt. This information can later be used to open the right image datablock in the Blender image editor when recalling the promt * Improvement: Create a hash for each image created and store that hash in the promt history as well. When recalling a prompt, load the image in the image editor if available * Use hashlib instead of hash() Co-authored-by: Carson Katri <Carson.katri@gmail.com> * Fix: Need to import hashlib before we can use it Co-authored-by: Carson Katri <Carson.katri@gmail.com>
…in the image editor when there is an image with the same name as the seed value in the promt.
When there is an image datablock with the same name as the seed from the promt to be recalled, open it in the image editor. When there is more than one datablock with the seed in it's name (like 12345.001, 12345.002 etc.) do not change the image open in the image editor and give a notification to the user.
image_sync_with_promt_history.mp4