Releases: Acly/krita-ai-diffusion
Version 1.32.0
Download krita_ai_diffusion-1.32.0.zip
Preview and Apply options
This release allows some additional customization for what happens when images finish generating and are applied.
- Finished Generation - Choose an action to take when a new result becomes available: do nothing, preview (default), or apply right away.
- Apply Behavior - Choose how results are applied to the canvas. The second drop-down is for region-specific behavior.
- no regions: Choose to modify the active layer, or create a new layer either on top or above the active layer.
- with regions: Choose how to modify or group region layers. Alternatively don't split results into region layers and follow the non-region behavior instead.
You can find the new options in Interface settings.
Other Changes
- Support alpha channel of image layers in custom workflows (Krita Image Layer node)
- Added Catalan translation by @vmalonso
- Fixed multi-selection for generated images #1641
- Resize inputs for Reference and Face control layers instead of crop to make them squares #1648
- Added option to force tiled VAE decoding in Performance settings #1650
- Added Laplace scheduler to custom scheduler options #1628
- Set conditioning to zero if there is no prompt (reduces noise for SDXL and some other models) #1628
- Improved ComfyUI server detection when starting the Plugin for the first time
- Improved report of missing models in connection settings
- Added network timeout for requests to ComfyUI server
- Managed server: upgraded PyTorch & ComfyUI versions
Contributors
Assets 3
Version 1.31.2
Download krita_ai_diffusion-1.31.2.zip
Changes
- Fixed history scroll to top when removing result images (also triggered when reaching maximum history size) #1600
- Result images are no longer cropped by default when they are applied #1598 #1608
- This allows to use custom workflows for upscaling by returning images larger than the canvas
- The "result applied" status (⭐ icon in history) is now persisted in documents #1613
- When there is a preview layer at the time a document is saved, it will now be reused after loading
- Fixed auto-completion when text contains >2 byte wide unicode symbols (eg. emojis) #1592
- Fixed input masks for custom workflows with resolution that aren't 4-byte-aligned #1602
- Fixed the root prompt being duplicated when a single region covers the entire image #1617
- Fixed a crash when opening more than one document with an embedded custom workflow
- Fixed upscale button remaining disabled after an error occured
Version 1.31.1
Download krita_ai_diffusion-1.31.1.zip
Changes
- Fixed error when using "prompt (negative)" parameter in custom workflows #1579
- Fixed crash when using Shift+Enter in prompt text box to start generation #1580
- Fixed error when debugpy is installed in a local Python 3.10 environment #1582
- Restrict the model architecture selection in Style settings to compatible models (anything else just leads to errors during generation)
- Automatically disable dynamic caching when using Flux Fill model (can cause garbage output)
Version 1.31.0
Download krita_ai_diffusion-1.31.0.zip
Custom Workflow Animations
The Custom Graph workspace now supports animation workflows which interact with Krita's animation timeline. Keyframes can be used as input and output, and will be available as image batches in ComfyUI. Documentation
CustomAnimation.mp4
Note
This requires a custom ComfyUI workflow!
Resizeable Prompt Box
For all those with entire pages worth of super secret prompt sauce.
Implementation courtesy of @FeepingCreature
Tiled Upscale Changes
New/changed parameters: #1117 #1272
- Image Guidance: this still controls the strength of the "Unblur/Tile" control-net, but can now be set to any %
- Tile Overlap: can now be set manually. Larger overlap will reduce visible seams between tiles, but increase generation time. Recommended to leave it on "Automatic" unless you see seam artifacts.
Quick Resolution
The resolution at which images are generated can now be adjusted quickly in the queue button dropdown. This has the same effect as the Resolution multiplier in the performance settings, and will override it when set.
The generate button now also indicates when the "Fixed seed" or "Resolution multiplier" settings are modified.
Other Changes
- Added a switch to toggle LoRA on/off in the style settings #1568 by @moonwhaler
- Support for dynamic caching (first block cache) via Comfy-WaveSpeed nodes
- Improves performance, especially for Flux.
- Must be enabled in Performance settings, and may degrade quality slightly.
- Can now drag-resize the parameter area in the custom graph workspace
- Fixed insightface install on windows not using a pre-built package #1553
- Fixed inpaint resolution issue when using 2-pass generation with control layers active
- Fixed inpaint resize/crop mismatch when using maximum pixel count or resolution multiplier
- Improved error messages for missing model files (fixes "StopIteration" error)
- Add info about supported workloads and models after connecting to custom ComfyUI server
- Allow using TAB key for auto-completion #1525
- Fixed remaining prompt headers in history after discarding results #1505
- Don't run ComfyUI in forced UTF-8 mode - this can fix some encoding issues but may introduce others...
- Fixed bug where changing settings no longer worked after an unrelated setting had been changed
- The download_files.py script can now download models in parallel
- Added Python 3.12 to prioritized versions when creating a managed install environment
- Update websockets library to version 14.2
Contributors
Assets 3
Version 1.30.0
Download krita_ai_diffusion-1.30.0.zip
Illustrious / NoobAI
Illustrious XL is a diffusion model trained extensively on (mostly anime-themed) illustration content. NoobAI XL adds further training on top. Both are based on the SDXL architecture, but weights have diverged significantly enough to require their own set of control models.
This release adds a separate "Illustrious" workload:
- NoobAI Control models (ControlNet, IP-Adapter) are used for styles using Illustrious architecture
- Managed install offers checkpoint and control models as download option
- Built-in style "Anime (Noob AI XL)" with recommended settings
- Supports both epsilon and v-prediction
- Seamless fill is disabled by default (there is no inpaint model so far)
Important
Most Illustrious checkpoints are detected as "SDXL" because they cannot be distinguished. You can configure
the correct architecture manually in Style settings > Checkpoint configuration (advanced) > Diffusion Architecture
The v-prediction models are detected automatically (no configuration needed).
User Guide
The documentation that was previously on the Wiki has moved to a new place: docs.interstice.cloud
There is a lot of new content too:
Something that was frequently requested but not easily possible on the Wiki was a full list of models recognized by the plugin: Model Database
Regional LoRA
LoRA which are added to text prompts of regions are now applied to that region only. Previously LoRA could only be applied to the entire image.
The initial implementation either requires a lot of VRAM or runs very slowly! Only use it if you really need it.
Other Changes
- Improved compatibility with custom tag lists (for auto-completion) by @BetaDoggo
- Lowered the default minimum resolution for SDXL to avoid resizing where it isn't required #1470
- Style settings are now locked when a built-in style is selected (with an option to make a copy to edit it)
- Made CFG rescale for v-prediction models configurable in style.json files (no UI for now)
- Reorganized model packages for the managed server install to be grouped by base model
- Upgraded managed install to PyTorch 2.5.1
- Fixed an assertion when there is a node with a "type" parameter in a custom workflow #1473
- Fixed mismatch between image and mask size when using resolution multiplier #1532
- Fixed an infinite recursion in custom graph workspace which could lead to hangs and broken UI when opening documents
- Fixed issues when active layer change is emitted before the layer is part of the node tree #1475
- Fixed empty error box being shown after a connection error is resolved
- Fixed long layer names stretching the animation workspace UI
- Moved model database to a separate JSON file (resource paths are still in resources.py)
- Added style models (eg. Flux depth/canny) to extra_model_paths
Contributors
Assets 3
Version 1.29.0
Download krita_ai_diffusion-1.29.0.zip
Flux Updates
"Reference" control layers now work with Flux. They use the Redux model to add content from one or more image layers to the generation.
There is also limited support for the Depth and Canny Loras released by Black Forest Labs. You can use the "Fill" model to Fill/Expand/Add/Remove content, but it's not fully integrated for now and must be selected manually.
See Flux Support for instructions. Keep in mind this is intended for those who want to experiment, expect changes in the future.
Other Changes
- Added "grace period" between making changes to the canvas or parameters, and starting a live generation. by @modelflat
- If generation is very fast, it will start immediately (as before)
- If it takes a bit longer there will be a tiny delay to wait for additional changes
- Will always schedule generation regularly after some time even when constantly editing the image
- Added context menu entry to copy image generation metadata to the clipboard #1421 by @mmuellerdm
- Added detection for some non-Flux GGUF models (eg. SD3.5 medium/large)
- Custom workflows: text prompt input box now respects the line count setting #1416 by @mmuellerdm
- Custom workflows: added context menu entry to copy generation parameters of previous results by @ronan36880
- Custom workflows: reserve more space for parameters in case none of the groups are expanded #1420
- Fixed Lora strength not being displayed correctly in Style settings in some cases
- Fixed cancel of active image generation when using cloud
- Open auto-complete only if there are at least 3 characters excluding trailing space #1417
Contributors
Assets 3
Version 1.28.1
Download krita_ai_diffusion-1.28.1.zip
Changes
- Fixed "missing parameter" error when using groups or explicit order in custom workflows #1393 #1414
- Fixed initial value for style widgets in custom workflows #1393
- Add
unetandtext_encodersfolders to built-in extra model paths #1395 - Update upscale button text based on context #1269
- Fixed broken link on welcome page #1388
- Improve error message when custom node pack is missing
Version 1.28.0
Download krita_ai_diffusion-1.28.0.zip
Built-in Style Model Updates
The built-in styles now refer to a range of models rather than one specific checkpoint, which allows for smoother updates. The following changes have been made to the recommended checkpoint list:
- [SD1.5] Serenity 2.1 (may replace RealisticVision5)
- [SDXL] RealVisXL 5.0 (may replace Juggernaut6)
- [SDXL] Pixelwave 11
As before those are just recommendations to get started with, you can always make your own choice.
New additions are also available on interstice.cloud
Custom Workflow Parameters
- Can now explicitly define the order of parameters
- Added parameter groups to aggregate related inputs
- Added button to reset all parameters in a group
Please check the documentation on how it works!
Managed Server Model Location
Models are now stored next to the ComfyUI installation folder, to allow smoother updates.
Eg. from .../ServerPath/ComfyUI/models to .../ServerPath/models.
Migration should happen automatically. Please find and put your models there afterwards (or use extra_model_paths.yaml).
Other Changes
- Updated managed ComfyUI & GGUF nodes (enables SD3.5 medium, #1383)
- Disable upscale button while upscale is in progress #1270
- Upscale: only show target size on hover to avoid showing old information #1264
- Fixed small but constant memory leak (even when doing nothing), #1359 by @FeepingCreature
- Fixed custom workflow parameter node min/max not accepting negative values #1381
- Error messages are now confined to a box, added "Copy to clipboard" button #1351
- Fixed clamp not being applied to grow/feather values
- Fixed Lora selection box expanding the Settings window #1343
- Fixed auto-completion with trailing space #1294
- Fixed text weight editing across multiple lines #1379
- Automatically switch to remote custom workflow after clicking "Open Web UI" button
Contributors
Assets 3
Version 1.27.1
Download krita_ai_diffusion-1.27.1.zip
Changes
- Fixed issues with intermediate diffusion results showing up among the output images #1330
- Fixed Parameter node not being connectable to other inputs Acly/comfyui-tooling-nodes#23
- Fixed Parameter node defaults being lost on reload Acly/comfyui-tooling-nodes#25
- Fixed Custom Workflow UI reserving space for text output when there is none
- Use a shorter request timeout for automatic update checks at startup #1339
- Added scrollbar for long custom parameter lists #1340
Version 1.27.0
Download krita_ai_diffusion-1.27.0.zip
Stable Diffusion 3.5
- SD3.5 checkpoints can be used with the plugin
- Added support for T5 text encoder for SD3
- Managed install now also supports fp8 scaled checkpoint version
As with all new models, a lot of features don't work yet, but feel free to give it a try!
See #1328 for more information on where to get the model.
Custom Workflows: Send Text
You can now send text from your custom workflow to Krita. Who knows, it could be useful.
