JWildfireSwan · build 19625923 · published
I have now completed the last major feature I had planned for version 1.0: a variant that generates image content with ai txt2img (using the automatic1111 integration) that can be used as basic shapes in fractals.
Because the ai-generation takes a significant amount of time it works asynchronous and in multiple passes. In the first passes you will see a placeholder image, which then will be replaced "on the fly" as soon as the image-generation has completed.
It was really tricky to integrate this into the render pipeline.
There is also a new "ai txt2img"-random-flame-generator which utilizes the "ai_txt2img_rect"-variation to generate random fractals which contain random ai-generated base-shapes/images. It generates both a random fractal and a random prompt for feeding the ai.
And, there is really a lot of other stuff, see the full changelog for details:
VERSION 0.114.1239 (17.08.2025):
- new "ai_txt2img_rect"-variation which creates a square shape using the colors
of an ai-generated image by using the automatic1111-integration
- because the ai-generation taks a significant amount of time it works
asynchronous and in multiple passes.
- in the first passes you will see a placeholder image, which then will be
replaced "on the fly" as soon as the image-generation has completed
- there can only be one image generated by ai, at the same time. When you load
multiple flames using this variation, they will all show the placeholder
image until you start editing a particular flame
- play with the parameter random seed to try different images with the
same prompt
- new "ai txt2img"-random-flame-generator which utilizes the "ai_txt2img_rect"-variation
to generate random fractals which contain random ai-generated base-shapes/images
- it generates both a random fractal and a random prompt for feeding the ai
- it contains several specialized styles (currently: 5) and one more generic style
- main-editor: made the main buttons resizable so that the right toolbar does
not disappear too easily on small screen resolutions (SWAN-116, SWAN-119)
- detect screen size at startup and shrink UI if necessary (SWAN-115)
- flora-variation: removed leaf type 25 which worked too different to all the other leaf types
- corrected default 4k render resolutions to 3840 x 2160 and 2160 x 3840 (SWAN-117)
- added variation descriptions for atan, atan2_spirals, attractor_flow, busybrad,
de_stijl, drunken_tiles, flora, maelstrom, mandelbox2D, meeple, multi_ifs,
natural_foam, shape_warp, square_rand, tessellated-t, tess_shape
- variation-browser: made url's in a variation description "clickable"
- application-settings: new option "enable ai-features" to enable/disable
all ai-related features in the software. There are some users who do not
use ai-features at all, so they are not distracted by related controls
they never use. This optino is per default disabled, so ai-related features
are not shown by default.
- added a hint about the new option "enable ai-features" in the users manual
- flame-editor: camera reset to default now resets all values (SWAN-109)
- fixed: reset to defaults didn't reset displayed values (also applied to
flipping xforms) (SWAN-108)
- added improved support for multi-monitor-setups (SWAN-121)
- fixed: entered numbers were sometimes rounded (SWAN-120)
- added 27 presets to the "ai_txt2img_rect"-variation
- "ai txt2img"-random-flame-generator lso makes use of the presets of
the "ai_txt2img_rect"-variation (and generates random prompts)
- new setting "keep_screen_on" in the application settings, which controls
if Godot tries to let the display always on, i. e. blocks a screen saver.
This setting is enabled by default (SWAN-129)
- main-editor: double-click at a node in the flame-tree collapsed all nodes
below this node. This helps to keep an overview when editing flames with
deep structure (SWAN-82)
- improved the size and layout of the confirmation dialog
- fixed: indo did not undo selecting a new variation (SWAN-124)
- fixed: canceling variation browse resets params to defaults (SWAN-122)
- main-editor: made the gradient-controls a little bit smaller so that
they fit better and the "generate"-button stays inside the view
- triangle-editor: you may also hold <Alt> to snap to grid (SWAN-130)