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
Added for Insider+ supporters (⚡ icon in Civitai Browser). Utilizes an optimized connection for dramatically faster browsing of online model repositories.
15
-
-**Inference Enhancements:**
16
-
- Added **Prompt Amplifier**: click the magic wand 🪄 in the prompt editor to expand and enrich your ideas with selectable “Feel”.
17
-
- Added **HiDream** support (see [ComfyUI Examples](https://comfyanonymous.github.io/ComfyUI_examples/hidream/) for details).
18
-
- Added toggle for filtering Extra Networks by base model.
19
-
- Added Abstract Syntax Tree (AST) parsing for prompts for more robust internal structure understanding.
20
-
- Added hotkey (`Ctrl+Up` / `Ctrl+Down`) in prompt editors to adjust token/selection weight emphasis.
21
-
- Added **Plasma Noise** addon for text-to-image workflows.
22
-
- Added **Wan 2.1** Text-to-Video and Text-to-Image project types.
23
-
- Added **Align Your Steps** scheduler.
24
-
- Added support for **wildcards** in prompts (e.g., `{blue|green|red}`).
25
-
- Added **Rescale CFG** addon.
26
-
- Added **Swap Dimensions** button between width/height inputs.
27
-
- Added `Ctrl+Tab` / `Ctrl+Shift+Tab` shortcuts for navigating between Inference tabs.
28
-
-**Model Management & Browsing:**
29
-
- Added **Safetensor Metadata viewer** to the Checkpoint Manager context menu (thanks @genteure!).
30
-
- Added Wan 2.1 files to the HuggingFace model browser.
31
-
- Added **OpenModelDB** tab to the Model Browser.
32
-
- Added new autocomplete tag source: [Danbooru/e621 merged tags](https://civitai.com/models/950325?modelVersionId=1419692).
33
-
-**Package & System:**
34
-
- Added new Package Command for installing **Triton & SageAttention** in ComfyUI.
35
-
-**UI & UX:**
36
-
- Added Undo/Redo commands to text editor context menus.
37
-
38
-
### Changed
39
-
-**Inference:**
40
-
- Merged Inference GGUF workflows into the UNet model loader option (no longer a separate choice).
41
-
- FaceDetailers now inherit the primary sampler/scheduler/etc. by default (can be manually overridden via ⚙️ button).
42
-
- Slightly rearranged the FaceDetailer card layout.
43
-
- Inference "Extra Networks" selector now filters extra networks based on the selected base model.
44
-
- Improved the quality of Inference inpainting via underlying workflow upgrades.
45
-
-**Model Management & Browsing:**
46
-
- Updated the Civitai Model Browser base model selector to match the new Checkpoint Manager filter UI.
47
-
- Renamed some shared model folders to better reflect their contents.
48
-
- Performance optimizations for Checkpoint Manager (progress indicators now fully use Compiled Bindings).
- Added new Package Command (in the 3-dots menu) for installing Triton & SageAttention in ComfyUI
14
+
#### Inference Features
15
+
- Added Prompt Amplifier to Inference - click the magic wand 🪄 in the prompt editor to expand and enrich your ideas. Tailor the vibe with the ‘Feel’ selector and watch as your generations come to life with extra detail, coherence, and flair!
16
+
- Added support for HiDream in Inference - see [ComfyUI Examples](https://comfyanonymous.github.io/ComfyUI_examples/hidream/) for more details
17
+
- Added toggle for filtering Inference Extra Networks by base model
18
+
- Added Rescale CFG addon to Inference
19
+
- Added Swap Dimensions button between the width/height input in Inference
20
+
- Added Ctrl+Tab/Ctrl+Shift+Tab shortcuts for navigating between Inference tabs
21
+
- Added Align Your Steps scheduler to Inference
22
+
- Added wildcards to Inference prompts, e.g. `{blue|green|red}` will randomly select one of the colors
23
+
- Added Wan 2.1 Text to Video and Text to Image project types for Inference
24
+
- Added new autocomplete tag source to Inference - [Danbooru/e621 merged tags](https://civitai.com/models/950325?modelVersionId=1419692)
25
+
- Added Abstract Syntax Tree (AST) parsing for Inference prompts. This provides a more robust internal understanding of prompt structure, paving the way for future enhancements.
26
+
- Added hotkey (`Ctrl+Up`/`Ctrl+Down`) in Inference prompt editors to adjust the weight emphasis of the token under the caret or the currently selected text.
27
+
- This automatically wraps the token/selection in parentheses `()` if it's not already weighted.
28
+
- It modifies existing weights within parentheses or adds weights if none exist (e.g. `(word:1.1)`).
- Added Plasma Noise addon to Inference for text to image workflows
31
+
#### Model Management & Discovery
32
+
- Added Accelerated Model Discovery (Beta) (⚡ icon in Civitai Browser) for Insider+ supporters. Utilizes an optimized connection for dramatically faster, more responsive browsing of online model repositories.
33
+
- Added OpenModelDB tab to the Model Browser
34
+
- Added Wan 2.1 files to the HuggingFace model browser
35
+
#### User Interface & Experience (UI/UX)
36
+
- Added Undo/Redo commands to text editor context menus
37
+
#### Internal / Developer Changes
38
+
- (Internal) Introduced unified strategy pattern (`IConfigSharingStrategy`) to for handling different config file formats (JSON, YAML, FDS).
39
+
- Added support for configuring nested paths in JSON and YAML files (e.g. `paths.models.vae`) via dot-notation in `SharedFolderLayoutRule.ConfigDocumentPaths`.
40
+
- Packages can now use the `SharedFolderLayout` property to define a `ConfigFileType` and `ConfigSharingOptions` (like `RootKey`), without needing to implement custom configuration logic.
41
+
42
+
### Changed
43
+
#### Inference Features
44
+
- Improved the quality of Inference inpainting by upgrading the workflow behind the scenes. The workflow remains the same for you — just better results!
45
+
- FaceDetailers in Inference will now inherit the primary sampler/scheduler/etc. by default. You can still manually set these by enabling the options via the ⚙️ button on the FaceDetailer card
46
+
- Slightly rearranged the FaceDetailer card layout due to the above change
47
+
- Inference "Extra Networks" selector now filters extra networks based on the selected base model
48
+
- Merged Inference GGUF workflows into the UNet model loader option (no longer need to choose GGUF separately)
49
+
#### Model Management & Discovery
50
+
- Changed the names of some of the shared model folders to better reflect their contents
51
+
- Improved Checkpoint Manager memory usage (thanks to @FireGeek for the profiling assistance!)
52
+
- Performance optimizations for Checkpoint Manager (progress indicators now fully uses Compiled Bindings)
53
+
#### Package Management & Compatibility
54
+
- Upgraded HIP SDK installs to 6.2.4 for ComfyUI-Zluda and AMDGPU-Forge
55
+
- Updated install for kohya_ss to support RTX 5000-series GPUs
56
+
#### User Interface & Experience (UI/UX)
57
+
- Improved window state handling
58
+
- Updated some date strings to take into account the user's locale
59
+
#### Localization
60
+
- Updated Japanese, Brazilian Portuguese, Chinese, and Russian translations
61
+
#### Internal / Developer Changes
62
+
- (Internal) Upgraded FluentAvalonia to 2.3.0
63
+
- (Internal) Refactored configuration-based shared folder logic: Centralized handling into `SharedFoldersConfigHelper` and format-specific strategies, removing custom file I/O logic from individual package classes for improved consistency and maintainability.
64
+
- Migrated packages ComfyUI (incl. Zluda), VladAutomatic (SD.Next), Sdfx, and StableSwarm to use the unified system for configuration and symlink based sharing.
- Fixed RTX 5000-series GPU detection in certain cases
69
+
- Fixed Package Updates and Change Version not using stored PyTorch index and instead using the default recommended index.
70
+
- Fixed ComfyUI-Zluda not being recognized as an option for Inference or SwarmUI (for real this time)
71
+
- Fixed errors from invalid pip specifiers in requirements files
72
+
#### User Interface & Experience (UI/UX)
73
+
- Fixed Image Viewer animation loader keeping file handles open, which resolves 2 different issues (OSes are fun):
74
+
- (macOS) Fixed `FileNotFoundException` crash when navigating to Inference tab after deleting a Webp animation file previously opened in the Image Viewer Dialog.
75
+
- (Windows) Fixed Webp animation files unable to be deleted without closing the app first.
76
+
- Fixed Image Viewer `FileNotFoundException` on fetching image size, if navigating to a deleted image file.
- Fixed window disappearing on macOS when the saved window size is very small
79
+
- Fixed large white boxes appearing when tooltips are visible on macOS/Linux
80
+
- Fixed package images sometimes showing as blank due to concurrent image caching. Requests to same image resources are now de-duplicated
81
+
- Reduced memory usage from `ShowDisabledTooltipExtension`
82
+
#### Inference & Workflows
83
+
- Fixed some cases of missing custom nodes in SwarmUI
84
+
- Fixed Inference ControlNet Preprocessors using incorrect resolution and increased maximum of smallest dimension to 16384
85
+
- Fixed Inference Extra Networks card not allowing for more than one model at a time
86
+
#### Model Management & Discovery
87
+
- Fixed missing base model options in the Metadata Editor
88
+
- Fixed some crashes when using Accelerated Model Discovery
81
89
### Supporters
82
90
#### Visionaries
83
91
Our heartfelt gratitude goes out to our amazing Visionary-tier Patrons: **Waterclouds**, **Corey T**, **bluepopsicle**, **Bob S**, **Akiro_Senkai**, and **Ibixat**! Your exceptional support is fundamental to the ongoing development and success of Stability Matrix. We are immensely grateful for your partnership and belief in the project! 🙏
0 commit comments