Releases: Acly/krita-ai-diffusion
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.
Other Changes
Version 1.26.0
Download krita_ai_diffusion-1.26.0.zip
Custom Workflows
You can now run custom ComfyUI node graphs from within Krita.
Watch Video (YouTube) | Documentation & Guide
Features
- New "Graph" workspace dedicated to custom workflows
- Import workflow files (ComfyUI JSON API format) and run them
- Connect to ComfyUI running in the browser for synchronized editing
- Expose important parameters in Krita Plugin's UI for easy access
- Supports queued and live (update on change) generation
- Exchange canvas, image and mask layers, selections, style settings...
Important
Custom workflows work with managed or custom (local/remote) server.
Make sure to update comfyui-tooling-nodes if you're running a custom install.
Other Changes
- Expand LoRA trigger words directly inside the prompt when using auto-completion
- Trigger words are not added when the LoRA is added via prompt behind the scenes anymore!
- No change for LoRA added from the Style settings (trigger words are appended to the prompt)
- Fixed diffusion model of the first style in the list being reset to default #1227
- Fixed parenthesis in LoRA filenames being escaped when using auto-complete #1213
- Fixed handling of jobs which finish without generating any images #1182
- Added list of used style LoRA to image metadata (shown in tooltip)
- Support for Flux unblur control layer (jasperai/Flux.1-dev-Controlnet-Upscaler)
- Fixed color of progress bar after a model upload
- Show status message when (temporarily) disconnected from the server
- Improved error handling when a mensaged server process has stopped or crashed
- Improved error message when failing to read transferred images
- Added indicator when number of control layers is too high (Cloud)
- Fixed issues with open pose vector generation
Version 1.25.0
Download krita_ai_diffusion-1.25.0.zip
Automatic Updates
Future versions of the plugin (excluding this one!) can be downloaded and installed automatically.
By default the plugin will check for new versions when Krita is started. This can be disabled in the new "Plugin" settings tab.
Flux Inpainting
The inpainting model by AlimamaCreative is now used for Fill, Expand, and related modes.
It is an alpha version, results so far are mixed, but better than not having an inpaint model. See Flux Support for more info on installation and setup.
Other Changes
- The Guidance slider in Style settings can now be used to influence Flux guidance (default: 3.5)
- Fixed issues with installation of insightface on Windows (required for "Face" control layers) #1186
- Fixed issue with missing SSL (TLS) certificates on macOS #1181 by @voronin-ilya
- Support system certificate stores for SSL (TLS) in download_models.py script by @Rotonen
- Fixed mixup of Mistoline ControlNet for Flux and SDXL #1200
- Fixed model search path prioritization
Contributors
Assets 3
Version 1.24.0
Download krita_ai_diffusion-1.24.0.zip
Flux Extensions
- Support loading diffusion models, not just full checkpoints
- Support quantized models in GGUF format
- Support for Control layers: Line Art, Soft Edge, Canny, Depth
Flux integration is ongoing and things will change. Please read: Current state, Download & Installation
Other Changes
- Selecting a folder to filter LoRA no longer changes existing entries #1138
- Fixed LoRA not being sorted alphabetically #1139
- Improved error messages when selecting incompatible LoRA
- Added "Copy Style" option to the context menu for generated images #1040
- Show more information about how images were generated in the history #1172
- Limit the number of tiles when upscaling to very large resolutions
- Allow setting a proxy via environment variables for the download_models.py script #1160
- Fixed issues with installed numpy version #1135
- Updated to latest PyTorch (2.4.1, CUDA 12.4)
🥨 Join us on Discord!
Share, discuss, support or just chat: discord.gg/pWyzHfHHhU
Version 1.23.0
Download krita_ai_diffusion-1.23.0.zip
LoRA Model Upload
The new Upload button allows you to select a .safetensors file anywhere on your system and use it as a LoRA.
This works across all connections:
- Copies files to the appropriate place for local installs
- Uploads files to remote ComfyUI servers
- Interstice.cloud service now supports custom LoRA!
- During BETA: model size is limited to 300 MB (to be increased later)
- Currently no other limits, and no additional charges
Files are uploaded when needed, and cached on the server.
LoRA Trigger Words
Additional settings are now stored for each LoRA, and can be used across different Styles:
- Trigger words are added to the prompt whenever a LoRA is used
- The default strength can be set individually
This also works when adding LoRA via the prompt (eg. <lora:Ink scenery>).
Translations
- Thai by @hotfreshmilk
- Korean by @sejoung
Thanks to all who contributed translation updates!
Other Changes
- Added button to copy / duplicate an existing Style #1084
- Fixed performance issue for small Fill regions in high resolution images
- Fixed error when there is a small selection with a very large context
- Fixed error when using high strength Refine region with SD3 or Flux
- Remember last used translation toggle state #1087
- Ignore Clip Skip and other base model settings when they are not supported (eg. SD3, Flux)
- Ignore control layers which aren't supported by the model or server (UI still shows a notice)
- Removed LCM and Lightning sampler presets from "official" list (existing Styles can still use them)
- Allow to continue an incomplete server installation in more situations instead of reporting an error
- Improved error messages for common connection issues
- Improved error message when trying to send really large images (>16k)
Important
Remember to update ComfyUI if you are using a custom install!
Latest version of comfyui-tooling-nodes and comfyui-inpaint-nodes are required.
Contributors
Assets 3
Version 1.22.0
Download krita_ai_diffusion-1.22.0.zip
Translations
Thanks to everybody who contributed translations!
- Chinese (simplified) by @syl1130
- Spanish by @flaviohr
- Japanese by @picolet21
- Brazilian Portuguese by @flaviohr
- Russian by @Drakosha405
- Chinese (traditional) by @PeterDaveHello
- Turkish by @alperktt
- French by @JonesJugHead
Tag Auto-Completion
You can now enable auto-completion for popular tags in Interface Settings. If you frequently use checkpoints that have been trained on tags, this allows you to compose precise prompts. Implementation by @FeepingCreature
Usage hints:
- Only enable tag files which the model has been trained for!
- Most tags are focused around anime characters
- You can add your own tag lists
Flux support (experimental)
There is initial support for the Flux models from blackforestlabs.ai. Only basic generation, refine and masks are supported. Please use one of the following checkpoints:
- Flux.1 [dev] to
ComfyUI/models/checkpoints - Flux.1 [schnell] to
ComfyUI/models/checkpoints
Use the built-in preset to get started. Keep in mind this is a much larger model compared to even SDXL (recommended: 16+GB VRAM, 32 GB RAM). You can also try out Flux via cloud service.
Docker image changes
For those who self-host on services like runpod.io or vast.ai, the new templates will only download a subset of models from now on. The total size was getting too large. You can customize which models you would like to download, read more here or in #973 by @daniel-richter
The download_models.py script now offers packages similar to the UI installer, see the changed list of arguments here.
Other Changes
- Enabled control layers when generating animation frames
- Fixed invisible item in the custom inpaint context drop-down #975
- Fixed error when selecting a prompt text box for a region linked to an empty layer group #990
- Fixed error when running the same upscale job twice
- Fixed issues with prompt translation when there is no connection
- Indent advanced checkpoint settings
Contributors
Assets 3
Version 1.21.0
Download krita_ai_diffusion-1.21.0.zip
Localisation
Interface Language
This release adds support for switching between languages for the plugin UI. There aren't any translations yet, but the framework is there. If you are interested in providing a translation, please read Contributing Translations. Open an issue or Pull request to get it added to future versions of the plugin.
Text Prompt Translation
The second part of localisation is automatic translation of text prompts into English. Select the language you want to translate in the Interface settings to activate the feature. You will get an indicator in all text boxes and can now enter instructions in your language of choice.
You can click the indicator to toggle translation on/off. Usually the translation happens behind the scenes, but you can Ctrl + click to translate immediately and replace your text with the translated version. Read more on using prompt translation
Translation happens locally with a small AI model, which is downloaded automatically on first use.
Universal ControlNet for SDXL
This release supports the Xinsir Union ControlNet model for SDXL. It combines many control modes (scribbles, lines, depth, normals, pose, segmentation...) into one model.
This does not change how you use control layers! Simply pick the mode from the drop-down as usual.
Installing (Managed Server)
The model can be downloaded in the Connection settings. It is now the only default option for new installations.
If you have an existing installation, note that it will not automatically delete previous models. You will have to do it manually if you want to save space, or re-install.
Installing (Custom Server)
- Download the model from huggingface.co/xinsir
- Rename to
xinsir-controlnet-union-sdxl-1.0-promax.safetensorsand put it inComfyUI/models/controlnet/- (other names are fine as long as they contain union-sdxl)
- Other specific control models will still work, and take precedence if they exist. Union model is used as fallback whereever it applies.
As usual, make sure ComfyUI & custom nodes are updated!
NSFW Filter
There is now an optional NSFW filter. It helps to mostly avoid accidental naughty things in places where that's not appropriate.
The options are:
- Basic filters out most explicit sexual content, but allows some (partial) nudity and suggestive images
- Strict attempts to filter out anything remotely suggestive
Warning
The filter is not perfect, and some images may slip through. Please keep this in mind when considering scenarios where eg. children are involved.
Other Changes
- Pinned the pytorch version for managed servers to avoid breaking installation mid-release #966 #967
- Using pytorch 2.3.1 for now because a lot of people have missing DLL issues with 2.4.0
- Reduced default line art control layer strength for SDXL
- Added configurable keyboard shortcut for Live mode "Apply in new layer" #959
Version 1.20.1
Download krita_ai_diffusion-1.20.1.zip
Changes
- Fixed hang/crash when replacing layer content in Live mode #922
- Fixed crash when previewing generation results after using "Flatten Layers" operation #836
- Fixed issues with Fill Layer and applying Live results not working in some cases #928
- Fixed Ctrl+Backspace shortcut (remove previous word) not working in text prompt editor #798 (by @dyscorv)
- Added some alternative search paths for xinsir control models #935
- Clean up interrupted and unfinished jobs from job queue after server disconnect #814
- Fixed opening generated images from ComfyUI where prompt text comes from nodes #920
Contributors
Assets 3
Version 1.20.0
Download krita_ai_diffusion-1.20.0.zip
UI Improvements
- Improved performance during Live painting with Regions #883
- The strength slider and % value now snaps to full sampling steps (by @FeepingCreature)
- Previously was a fixed interval, where sometimes changes seemingly had no effect
- Now snaps to the previous/next value that would produce a change in the image
- Optionally displays the full step count (see Settings ⮞ Interface ⮞ Show Steps)
- When creating a new document, the most recently used parameters are used as defaults #902
- Affects things like seleted Style, Inpaint settings, and batch count
- When opening images generated by Auto1111 or basic ComfyUI workflows, the prompt is now read from the image
- Added context menu for discarding all generated images (clear history) #728
- Documented
{prompt}placeholder in Style settings #901
New Models
- Added xinsir depth-sdxl-1.0 SDXL depth control model
- Added xinsir tile-sdxl-1.0 SDXL unblur control model (new default)
- Added sdxl_segmentation_ade20k SDXL segmentation control model
Note: These are alternative choices to the previous models, which continue to work. There is no need to replace them.
New Schedulers
There are various new options for schedulers when creating custom sampler presets (thanks @Danamir!):
align_your_stepsgits(this is now used by the built-in "Fast" preset)polyexponential
Important
If you are using a custom ComfyUI installation, make sure to update it along with custom nodes.
Bugfixes
Contributors
Assets 3
Version 1.19.0
Download krita_ai_diffusion-1.19.0.zip
Model Updates
There have been a lot of new and improved models for SDXL lately. This update changes quite a few of the plugin's defaults, with some optional choices also added to the list.
SDXL ControlNet models
| Mode | 1st priority | 2nd priority | 3rd priority |
|---|---|---|---|
| Scribble | 🌟 XinsirScribble | 🌟👍 MistoLine | SAI-Sketch |
| Line Art | 🌟 XinsirScribble | 🌟👍 MistoLine | SAI-Sketch |
| Soft Edge | 🌟👍 MistoLine | 🌟 Xinsir-Scribble | |
| Canny Edge | 🌟 XinsirCannyV2 | 👍 SAI-Canny | 🌟 MistoLine |
| Pose | 🌟👍 XinsirOpenPose | ThibaudOpenPose | |
| Segmentation | 🌟 abovzv-ade20k | ||
| Stencil | 🌟 QR Code Monster |
🌟means it's a new model, 👍means it's the default download option for the integrated installer.
If you have multiple models installed the priority defines the order in which they are selected.
Sometimes the first choice isn't the default, usually because they're really close in terms of quality, but the other model is smaller (requires less space and loads faster). You may also notice that most of the line modes are covered by the same models, it works pretty well regardless!
ControlNet preprocessor models
Those are the models that turn your current image into lines, a depth map, a pose skeleton, etc. They apply to both SD1.5 and SDXL! There are three changes here:
- Soft Edge now uses AnyLine model. It will be downloaded automatically.
- Depth now uses version 2 of DepthAnything. It will be downloaded automatically.
- Canny Edge was changed to be more sensitive. No model/download required.
Realtime/Live models
Those are LoRA used to generate images significantly faster with lower step count. There is usually a trade-off in quality.
In this update Hyper-SD replaces LCM as the default choice for Live mode. This means the model is required to use the plugin. It will be downloaded when you update. Compared to LCM they are much more faithful to the model's usual output and retain most of the quality.
The LCM models are no longer required, but can still be used as before. For your custom Styles you can switch to Hyper easily by selecting the "Realtime - Hyper" sampler preset. One of the advantages is that you can tweak the CFG value now to get more defined results.
Custom ComfyUI installs
This section is for those who manage their own ComfyUI server. As usual, please make sure ComfyUI and custom nodes are up-to-date!
You will have to download the Hyper-SD 8-step CFG LoRA and put it into ComfyUI/models/loras:
For the new control models, please follow the links above. To use the Xinsir models rename them into any filename which contains xinsircanny, xinsiropenpose, xinsirscribble (case-insensitive). If you already downloaded them from civitai those filenames will match.
Stable Diffusion 3
There is basic support for the Stable Diffusion 3 (SD3) base model, so you can give it a try! "Basic" means that things like control layers and regions are not supported, and you need to download models manually from HF SD3 Medium. The required files are:
- sd3_medium.safetensors to
ComfyUI/models/checkpoints - clip_g.safetensors to
ComfyUI/models/clip - clip_l.safetensors to
ComfyUI/models/clip
Please use the (somewhat temporary) SD3 sampler preset!
Other Changes
- Added ZavyChromaXL to recommended checkpoints and made it the default for "Digital Artwork (XL)" style
- Fixed SSL issues when using a https connection on Mac #840 by @DrewWalkup
- Fixed control layers disappearing when unrelated layers are removed #823
- Fixed "/ComfyUI" being appended when selecting existing server installs #859
- Fixed the "Show negative prompt" toggle not having any immediate effect #857
- Fixed error being reported when creating Control Layers with active Filter Mask layers #846
- Fixed LoRA not appearing in prompt history (again) #855
- Paint layers are now converted into groups correctly also when applying multiple results from the same batch
- Properly track linking/unlinking regions - previously regions could end up linked to the wrong layer in saved .kra files
- Fixed crash in rare situation where the image is a tiny bit large than the preferred checkpoint resolution
- Fixed some issues with apply when replacing layer content
