v2.16.1

Added

• Added automatic text encoder and VAE selection to the Inference Model card. Selecting a model now fills any empty encoder slots and the default VAE with the matching local files for the detected workflow, so you don't need to know which files pair with which architecture (e.g. qwen_3_4b or qwen_3_8b + Flux.2 VAE for Flux.2 Klein, clip_l + t5xxl for Flux, qwen_3_06b + qwen_image_vae for Anima). Anything you pick manually is never overridden
• Added a misplaced-model warning to the Inference Model card with a one-click Move button. If a model sits in a folder that can't work with the selected workflow (like a Z-Image or Anima file in the StableDiffusion folder), a compact warning explains the problem instead of letting generation fail with a cryptic ComfyUI error. The Move button relocates the file and its metadata to the right folder, then re-selects it. Dismissible per model
• Added a native macOS menu bar with the standard application menu — About Stability Matrix, Settings… (⌘,), and the usual Services / Hide / Quit (⌘Q) items

Changed

• The Inference Workflow selector now switches the model loader to match the chosen profile, showing or hiding the separate encoder and VAE fields as appropriate. It will never switch to a loader that can't load the selected file; you get the warning above instead
• Renamed the "Anima / SD" workflow profile to "Anima". Anima has no all-in-one version, so it's now handled like Z-Image: standalone model in DiffusionModels with a separate text encoder and VAE
• Image Lab's Flux.2 Klein model checks now match the text encoder to your selected UNET variant (4B vs 9B), and switching variants updates the status banner immediately
• Image Lab's Flux Kontext and Qwen Image Edit providers now show ComfyUI's actual workflow rejection message instead of a generic "Generation failed" (Flux.2 Klein already did this)
• Image Lab now opens the same ComfyUI error detail dialog that Inference uses when a workflow is rejected or a node fails mid-generation, showing the full error JSON instead of a truncated toast
• Keyboard shortcuts now use ⌘ (Command) instead of Ctrl on macOS — save/open, Generate, undo/redo, copy/paste/cut, tab navigation, and the mask editor all follow the platform convention, and context menus show the ⌘ glyph

Fixed

• Fixed #1659 - Z-Image and Anima workflows hiding the Text Encoder selectors and passing an invalid None CLIP input to ComfyUI; standalone workflows now expose and automatically fill compatible text encoders and VAEs
• Fixed #1654 and #1658 - Inference Image-to-Image masks could disappear in Linux AppImage builds, fail to appear in the image-card preview, or crash the app when saving an enabled clipping mask because Avalonia retained pixels owned by a disposed Skia image; converted mask bitmaps now own their pixel data and previews refresh after editing, restoring a project, or loading image dimensions
• Fixed #1660 - CivArchive downloads always saving to the model folder root; the primary Download button now offers the inferred folder, existing subdirectories, and a custom folder picker, while filename patterns containing path separators create missing nested subfolders and keep downloads inside the selected models directory
• Fixed #1661 - Custom UNet workflows missing current ComfyUI encoder types such as sdxl; the selector now includes current CLIP loader values while retaining single-encoder compatibility and migrating legacy HiDream casing
• Fixed #1664 - Gemini failures with a saved key being misreported as "API key not configured"; Image Lab now shows the actual invalid-key, billing, quota, or permission error and explains that Nano Banana image generation requires a paid-tier API key from a Google AI project with billing enabled
• Fixed "No text encoders configured" errors when generating with an all-in-one checkpoint after a UNet model had been selected in the same tab
• Fixed Qwen Image Edit in Image Lab failing mid-generation when a wrong-size Qwen2.5-VL text encoder was installed. The 7B encoder is now required, and the correct download is offered when it's missing
• Fixed Image Lab reporting "all models present" for Flux.2 Klein 9B setups that only had the 4B text encoder (and vice versa). The matching encoder download is now offered
• Fixed Image Lab model and LoRA dropdowns hiding files whose CivitAI base model tag is unrecognized (commonly "Other"), even when the filename clearly matches the provider
• Fixed Animagine XL and other SDXL models with "anima" in the name being misdetected as the Anima architecture
• Fixed the running-package exit confirmation not appearing when quitting via ⌘Q, the macOS app menu, or the Dock — it previously only showed when closing the window directly, so those paths could tear down running packages without warning. All quit paths now prompt
• Fixed an empty strip of space appearing below the native window title bar on macOS and Linux; the Windows-only caption area (icon, title, and min/max/close buttons) is now collapsed on those platforms so content sits directly under the system title bar

Performance

• Optimized SKBitmap-to-WriteableBitmap conversion used throughout Inference and Image Lab; pixel data is now copied directly native-to-native (single bulk copy when strides match, per-row otherwise) instead of round-tripping every scanline through a temporary managed buffer, reducing allocations and GC pressure on full-resolution images

Supporters

🌟 Visionaries

A massive thank you to our brilliant Visionaries: Waterclouds, bluepopsicle, Ibixat, Droolguy, snotty, LG, whudunit, MrMxyzptlk12836, Psilocyfer18731, KalAbaddon, and moon_milky2843! There's a little of your support behind every fix and refinement in this update, and we're grateful for all of it. A warm welcome to our newest Visionary, cusalapapen1481; it's wonderful to have you with us! 💛

🚀 Pioneers

And an equally big thank you to our fantastic Pioneer crew, all familiar faces this time around: Szir777, [USA]TechDude, SinthCore, Jisuren, Tigon, jweg79, rwx14662, Hurbie53, ahnhj.al, drew.lukas, Tuskaruho, Cjloha, Alligator1907, Bitti, Ghislain G, CommissarGiygas16050, qob97515211, bastardofbethlehem, and Zombop! You keep showing up for us, and that steadiness is a big part of how this project keeps moving forward. Thank you, truly, every one of you. 💛

v2.16.0

Added

New Feature: 🧪 Image Lab - Conversational Image Generation for ComfyUI

• We've added a brand new conversational interface for image generation! Image Lab lets you iterate on images naturally through chat, rather than just one-off prompts.
  • Local-First Power: Native support for Flux Kontext, Qwen Image Edit, and the Apache 2.0-licensed Flux.2 Klein running entirely locally via your ComfyUI backend.
  • Smart Setup: Stability Matrix automatically detects and helps you download the specific models and LoRAs needed for these local workflows.
  • Interactive Tools: Drag-and-drop image inputs, use the built-in annotation tool to draw on images, and keep persistent conversation history.
  • Cloud Option: Includes optional support for Nano Banana (Gemini 3 Pro / 2.5) and Nano Banana 2 (Gemini 3.1 Flash) for users who want to leverage external reasoning models.
• Added Regional Prompting addon to Inference - paint detailed masks to apply different prompts, strengths, and settings to specific regions of your image
  • Multi-layer mask editor with Photoshop-style interface for managing layers with independent masks, prompts, colors, and opacity
  • Professional brush tools: freehand brush/eraser with pressure sensitivity, rectangle/ellipse shapes with fill/stroke modes, paint bucket flood fill
  • Brush feathering/softness control for smooth, blended mask edges (0 = hard edge, 1 = soft/blurred)
  • Per-layer prompt and strength controls, export/import masks as PNG, duplicate layers, image reference layers for tracing
  • GPU-accelerated rendering with compact gzip-compressed metadata serialization
• Added official Inference support for the Z-Image (Base + Turbo), Anima, and Flux.2 model architectures — workflow-appropriate text encoders, latent shapes, schedulers, and model sampling (AuraFlow for Z-Image, Flux2Scheduler for Flux.2) are wired up automatically across Text-to-Image and Image-to-Image
• Added an Inference Workflow selector to the Model card with profiles for Default/Checkpoint, Flux, Flux.2, Z-Image Base/Turbo, Anima, HiDream, and Custom
  • Auto (default) detects the workflow from the model's CivitAI metadata, with filename fallbacks for models without metadata, and shows the resolved profile inline below the selector
  • Sparkle button applies recommended sampler / scheduler / steps / CFG presets for the active workflow — e.g. res_multistep / simple / 8 steps / CFG 1 for Z-Image Turbo, er_sde / simple / 30 steps / CFG 4 for Anima, euler / 20 steps / CFG 5 for Flux.2
  • Choosing a non-Auto profile reveals a manual Encoder Type selector for advanced overrides (e.g. running Z-Image Turbo with the sd3 encoder)
  • Opening the model browser from the Model card pre-filters to the workflow's compatible base models, without overwriting your saved picker filters
• Added CivArchive model browser with details page, image viewer, version selector, trigger words, and in-app downloads with tracked progress
• Added a checkpoint organizer for previewing and reorganizing local models using connected metadata-driven folder and filename patterns (requested in #280, #424)
• Added a new Model Picker dialog for Inference with grid/list views, search, filtering, and NSFW overlay
• Added browse buttons to all model dropdowns in Inference (Model, Refiner, VAE, Text Encoders, CLIP Vision)
• Added an inline search box to model combo box dropdowns with fuzzy matching
• Added a Source button in the Inference SamplerCard that one-click matches your generation Width/Height to the loaded source image — available in Image-to-Image whenever a source image is selected
• Added popularity counts to booru-style tag completions in the prompt editor; descriptions now show entries like 12.3K · artist so the more common tags are easier to spot at a glance
• Added a settings gear button to the CivitAI browser's Base Models filter flyout that jumps straight to the base model filter configuration in Settings
• Added er_sde and res_multistep to the Inference sampler list
• Added stable_diffusion, flux2, and lumina2 Encoder Type options for UNet workflows
• Added a Bitsandbytes NF4 launch option to Stable Diffusion WebUI Forge - Neo for low-bit (--bnb) inference
• Added an Activity center: the sidebar download panel now has a Notifications tab alongside In Progress. Toasts are clickable — jumping to the downloaded folder, the originating page (e.g. Inference), or the activity panel — and persist into a session notification history (every notification is recorded, even ones suppressed by your settings) with read/unread indicators and a combined unread + active-download badge on the sidebar item
• Added an "Always Show Scrollbars" toggle under Settings → Appearance. Defaults on — vertical scrollbars stay visible at their full thickness and reserve real layout space instead of fading to a thin overlay-style bar that only thickens on hover. Toggle off to restore Avalonia's classic auto-hide behavior. Single-line numeric inputs (e.g. SamplerCard Width/Height) keep their auto-hide regardless so spin-buttons aren't followed by a phantom bar
• Added new shared model folder categories — Style Models, Audio Encoders, Model Patches, and Background Removal — for ComfyUI's style_models, audio_encoders, model_patches, and background_removal directories. Models in these folders are now indexed and symlinked alongside everything else (e.g. Flux Redux / B-Lora style models, audio encoders for video/audio workflows, BiRefNet background-removal models)
• Added Intel GPU support for ComfyUI
• Added "Run Python Command" option to the package card's 3-dots menu for running arbitrary Python code in the package's virtual environment
• Added a recoverable error dialog for UI thread exceptions, with option to continue instead of exiting
• Added enable/disable toggle for environment variables in Settings, allowing variables to be temporarily disabled without deleting them
• Greatly expanded native Windows ROCm (AMD GPU) support (#1629) — the GPU detection matrix now spans Vega / GCN5 (Vega 56/64, Radeon VII) through the entire RDNA1/2/3/3.5 lineup and into RDNA4 (RX 9070 / R9700), using TheRock ROCm Technical Preview PyTorch builds. ROCm install and launch now run through a shared helper, so the same Windows-native path is available to ComfyUI, SwarmUI (its Comfy backend), reForge, InvokeAI, and Wan2GP - thanks to @NeuralFault!
• Added optional ROCm package commands for ComfyUI on Windows — one-click install of SageAttention, Flash Attention, bitsandbytes, and the ROCm SDK devel module (for compiling extensions/modules against your installed ROCm) from the package's command menu - thanks to @NeuralFault!

Changed

• Promoted the Encoder Type selector in the Inference Model card out of Advanced Options up to the main card body, so it's visible whenever a non-Auto workflow profile is active (and always when Custom is selected)
• Tidied up the Inference SamplerCard dimensions section — Source/Presets actions are shown as labeled buttons below the dimension row
• The Inference checkpoint dropdown no longer resets its scroll position every time the model list refreshes. The refresh now applies a single combined (local + remote) diff to the underlying source cache, rather than first resetting to local-only and then re-adding remote entries — which previously caused the open dropdown to scroll back to the top
• Local model autocomplete in the prompt editor now uses substring matching instead of prefix-only — typing any part of a model's filename surfaces it, with names that start with your search still ranked first
• Single-encoder UNet workflows (Anima, Flux.2, Z-Image) now use the matching CLIPLoader instead of assuming Flux-style dual encoders
• The CivitAI model details page now collapses the preview-image area and shows a small "No preview images available" hint when a model has no images to display, letting the description card take the full vertical space instead of leaving a large empty region above it
• Improved the Gemini API error message in Image Lab when the API returns 401/403 to point users at Google's API key restriction policy (which starts blocking unrestricted keys on June 19 2026)
• Improved safetensor checkpoint classification to correctly detect UNet-only models for Wan Video, HiDream, Z-Image, Hunyuan3D, and diffusers-format Flux architectures, ensuring they are routed to the DiffusionModels folder
• GGUF checkpoint downloads now go directly to the DiffusionModels folder instead of StableDiffusion
• Updated AI-Toolkit to install torch 2.9.1 / torchvision 0.24.1 / torchaudio 2.9.1 from the cu128 index to match upstream (ostris/ai-toolkit), with a cu126 fallback for legacy NVIDIA GPUs; also pin numpy to 1.26.4 to avoid a numpy 2.x ABI break in scipy/diffusers that crashed training runs
• Pinned kohya_ss torch to 2.7.0 / torchvision 0.22.0 (cu128) to match upstream's requirements_pytorch_windows.txt instead of resolving an untested latest, keeping the cu126 legacy-GPU fallback
• Pinned reForge torch to 2.9.0 to match upstream (modules/launch_utils.py)
• Updated ComfyUI installs to cu130 (cu126 for legacy NVIDIA GPUs) / rocm7.2 torch indexes depending on GPU
• Upgraded the bundled Visual C++ redistributable from 2015–2019 (v16) to 2015–2022 (v17, build 14.40.33810+), required by modern native dependencies such as PyTorch and ONNX Runtime
• Video files can now be opened directly from the Output browser
• Videos will now appear with thumbnails in the Output browser
• Configured portable Git to suppress detached HEAD advice messages
• PyTorch TunableOp is now disabled by...

v2.15.8

v2.15.8

Added

• Added support for the civitai.red (mature-content) domain — NSFW CivitAI links now open and copy as civitai.red URLs, and pasting a civitai.red URL into the CivitAI model browser search works the same as a civitai.com URL

Changed

• The CivitAI base model type filter now uses CivitAI's official /api/v1/enums endpoint, with fallbacks to the previous technique and a built-in list, so the filter stays populated even if the CivitAI response format changes or the service is unreachable

Fixed

• Fixed #1608 - Crash when cdn fetch fails due to error notification not being shown on UI Thread - thanks to @NeuralFault!
• Fixed CivitAI model browsing breaking during Discovery API outages — the browser now falls back to the direct CivitAI API when Discovery returns a server error, authentication failure, or times out
• Fixed SwarmUI user settings (theme, output format, server configuration, etc.) and any user-added backend entries being overwritten when the install flow ran over an existing install — Settings.fds and Backends.fds are now merged with their existing contents instead of being rewritten from a stale template
• Fixed pip requirements handling for environment-marker dependencies - thanks to @NeuralFault!
• Fixed #1608 - Crash when cdn fetch fails due to error notification not being shown on UI Thread - thanks to @NeuralFault!
• Fixed ComfyUI-Zluda inheriting --enable-manager from the base ComfyUI launch options, which blocked the bundled custom-node manager from initializing - thanks to @NeuralFault!

Supporters

🌟 Visionaries

Heaps of gratitude to our Visionaries — Waterclouds, bluepopsicle, Ibixat, Droolguy, snotty, LG, and whudunit — for sticking with us release after release. Your encouragement, your patience while we chase down those last bugs, and the sheer fact of you being here keeps us showing up at the keyboard. We're so glad you're part of this little corner of the internet with us. And big warm welcomes again to our newest Visionaries MrMxyzptlk12836, Psilocyfer18731, KalAbaddon, RustCupcake, and moon_milky2843 — make yourselves at home, you're among friends! 💛

🚀 Pioneers

And the Pioneer crew — what a lineup. A massive thank-you to Szir777, [USA]TechDude, takyamtom, SinthCore, Commissar Lord Death, Ahmed S, SeraphOfSalem, and Jisuren — your steady presence, kind words, and patience as we've shifted things around mean more than you know. A heartfelt welcome back to Tigon, who's returned to the Pioneer ranks after a little time away — so glad you're back. 🎉 And a special hello to jweg79, who's been quietly supporting us for a while and just decided to step up and join the Pioneer crew this round — so happy to have you here. To our newest Pioneers, an enormous welcome: rwx14662, Hurbie53, ahnhj.al, drew.lukas, Firelight, joeto332987, Tuskaruho, Cjloha, Alligator1907, Bitti, damianpointdexter, and tmdcks! We're absolutely thrilled to have you with us. (And to our anonymous Pioneer out there too, our thanks reaches you — we see you. 💛)

v2.15.7

v2.15.7

This release is part of a larger story. Here's the short version:

Going independent

On March 25th, Patreon Trust & Safety removed the Stability Matrix page, under their policy against AI tools that can produce explicit imagery.

While we respect Patreon's right to govern their platform, banning us under this policy is exactly like banning a web browser because it can access NSFW sites, or banning VS Code because it can be used to write malware.

Our Stance

To be upfront: We believe software tools should not be restricted based on what users might hypothetically do with them. We will continue developing Stability Matrix exactly as it was meant to be – an open, unrestricted tool for the community.

The Community Response

Since the removal, the community's response has been incredible, and the message was clear: open-source tools should not be at the mercy of platform policies.

Building Something Better

That response is what made going independent possible. Stability Matrix is now directly supported by the people who use it, with no middlemen deciding what software is allowed to exist. Memberships at lykos.ai/membership, with significantly more of every contribution going into development.

Stability Matrix is free and open-source, and it always will be. That's not changing. Supporters fund development, servers, cross-platform builds, package compatibility work, and the people doing the work. If you believe in open-source tools that answer to their users, we'd love to have you with us.

Read the full story →

---

What's new in v2.15.7

PNG metadata is finally spec-compliant for CJK users. If you've ever generated images with Japanese, Chinese, Korean, or emoji in your prompts and had them come out as mojibake when opened in other tools, that's fixed. Non-Latin-1 content now uses proper iTXt chunks with UTF-8 encoding (#1535).

Settings files are now self-healing. If your settings file gets corrupted from an interrupted write, crash, or power loss, SM will now automatically recover from null bytes, truncated JSON, and missing brackets instead of crashing on startup (#1590). Settings saves are also atomic now, so the corruption is less likely in the first place.

Better Japanese text rendering in dialogs — bundled NotoSansJP is now used for CTextBlock and ContentDialog, fixing compressed/squished text and incorrect fonts in markdown dialogs.

A wave of download reliability fixes — including downloads from CivitAI/HuggingFace redirects (e.g. CivArchive), VPN connection failures, the "request message was already sent" retry bug, and the progress bar showing 100% immediately on fresh downloads (#1598).

Plus a bunch of other stuff:

Added

• Added single-instance window activation signaling so reopening the app restores and focuses the existing desktop window instead of launching a duplicate instance
• Added notification system with localizable banner and markdown detail dialog UI
• Added warning in data directory selector when an OneDrive folder is selected
• Added support in the Checkpoints page to distinguish standard updates from Early Access-only updates - thanks to @x0x0b!
• Added torch index for Strix/Gorgon Point Ryzen AI APUs on Windows - thanks to @NeuralFault!
• Added retry button to failed downloads - thanks to @NeuralFault!
• Added new Membership support in Account Settings with Patreon migration prompt

Changed

• Settings file saves are now atomic to prevent corruption from interrupted writes
• Updated torch indexes for A1111, ComfyUI, InvokeAI, and Forge-based UIs to rocm7.2 / cu128 depending on GPU - thanks to @NeuralFault!
• Replaced the "Become a Patron" footer button with "Support Us", linking to the new direct Lykos support page at lykos.ai/membership
• Updated the prompt dialog shown when enabling features like Accelerated Model Discovery to use Lykos accounts instead of Patreon linking
• Moved the Patreon connection in Account Settings to a new "Legacy Connections" section, only shown for users with an existing Patreon link
• Localized previously hardcoded strings on the Account Settings page (menu items, descriptions, section headers) and added Japanese, Korean, German, and French translations

Fixed

• Fixed an issue where Align Your Steps scheduler and Unet Loader workflows ignored Regional Prompting (and other addon) conditioning modifiers.
• Fixed the Package Manager "Add Package" teaching tip opening inopportunely while packages were still loading or after opening the add-package dialog
• Fixed bold text not rendering in markdown dialogs on Windows 11 due to Avalonia 11.3.x variable font regression with Segoe UI Variable Text
• Fixed Japanese text appearing compressed/squished in markdown dialogs by ensuring the bundled NotoSansJP font is used for CTextBlock rendering
• Fixed ContentDialog title and buttons not using the correct font for Japanese locale (NotoSansJP) when shown as overlay
• Added missing CBold and CItalic inline styles to the markdown style sheet
• Fixed downloads failing with "The request message was already sent" when the server doesn't return Content-Length on the first attempt, caused by reusing a consumed HttpRequestMessage in the retry loop
• Fixed downloads from sources that redirect to CivitAI/HuggingFace (e.g. CivArchive) failing with Unauthorized by resolving the redirect target URL and applying auth headers for the correct domain
• Fixed dropdown menu overlayed in Inference UI Model Cards not being scrollable on Linux - thanks to @NeuralFault!
• Fixed model downloads failing on VPN connections - thanks to @NeuralFault!
• Fixed #1598 - download progress bar showing 100% immediately for fresh downloads due to missing Content-Length fallback when Content-Range header is absent
• Fixed #1597 - reForge launch failing due to setuptools version
• Fixed #1596 - package installs and managed embedded Python startup being poisoned by inherited shell Python activation variables such as PYTHONHOME, PYTHONPATH, VIRTUAL_ENV, and Conda environment variables
• Fixed #1590 - Startup crash when settings file is corrupted. Settings files are now self-healing with automatic recovery from null bytes, truncated JSON, and missing brackets
• Potentially fixed #1578 - SocketException: Address already in use on Linux startup by cleaning stale interprocess socket files and reactivating the existing window
• Fixed #1397, #610 - duplicate pip package entries in results - thanks to @e-nord!

---

Supporters

🛡️ Going Independent

When Patreon removed our page on March 25th, we chose to stand by our software rather than change what it is. The first wave of supporters followed us to direct membership within hours — and we remember every single one of you who didn't wait for reassurance before stepping up. This next chapter exists because you built it with us.

🌟 Visionaries

Lifting up our legendary Visionaries: Waterclouds, JungleDragon, bluepopsicle, Bob S, and whudunit. Through every release, every milestone, and every twist of this wild journey, your support has been our north star. A huge welcome to our newest Visionaries Droolguy and snotty (leveling up from the Pioneer crew!), a warm welcome back to longtime Visionary Ibixat, and an equally huge welcome to LG and MrMxyzptlk12836, making their Stability Matrix debut straight at the Visionary tier - so glad to have you all on board! We cannot thank you enough for standing behind Stability Matrix!

🚀 Pioneers

And to our mighty Pioneer crew - the folks who keep the engine humming - thank you for everything! Massive shoutout to: Szir777, Noah M, [USA]TechDude, Thom, SeraphOfSalem, Desert Viber, Adam, ACTUALLY_the_Real_Willem_Dafoe, takyamtom, robek, Ghislain G, Phil R, Tundra Everquill, Andrew B, SinthCore, and Ahmed S. And a very warm welcome to our newest Pioneers Commissar Lord Death, Firelight, and jweg, plus a heartfelt shoutout to one more new Pioneer who joined us quietly through the Stripe migration - you know who you are, and we're so glad to have you!

And one more heartfelt thank you to everyone now supporting us directly through our new platform - this next chapter wouldn't be possible without your trust, and we're so grateful you've come along for the ride!

v2.15.6

v2.15.6

Added

• Added NVIDIA driver version warning when launching ComfyUI with CUDA 13.0 (cu130) and driver versions below 580.x
• Added legacy Python warning when launching InvokeAI installations using Python 3.10.11
• Added Tiled VAE Decode to the Inference video workflows - thanks to @NeuralFault!

Changed

• Disabled update checking for legacy InvokeAI installations using Python 3.10.11
• Hide rating stars in the Civitai browser page if no rating is available
• Updated uv to v0.9.30
• Updated PortableGit to v2.52.0.windows.1
• Updated Sage/Triton/Nunchaku installers to use GitHub API to fetch latest releases
• Updated ComfyUI installations and updates to automatically install ComfyUI Manager
• Updated gfx110X Windows ROCm nightly index - thanks to @NeuralFault!
• Updated ComfyUI-Zluda install to more closely match the author's intended installation method - thanks to @NeuralFault!
• Updated Forge Classic installs/updates to use the upstream install script for better version compatibility with torch/sage/triton/nunchaku
• Backslashes can now be escaped in Inference prompts via \\

Fixed

• Fixed parsing of escape sequences in Inference such as \\
• Fixed #1546, #1541 - "No module named 'pkg_resources'" error when installing Automatic1111/Forge/reForge packages
• Fixed #1545, #1518, #1513, #1488 - Forge Neo update breaking things
• Fixed #1529 - "Selected commit is null" error when installing packages and rate limited by GitHub
• Fixed #1525 - Crash after downloading a model
• Fixed #1523, #1499, #1494 - Automatic1111 using old stable diffusion repo
• Fixed #1505 - incorrect port argument for Wan2GP
• Possibly fix #1502 - English fonts not displaying correctly on Linux in Chinese environments
• Fixed #1476 - Incorrect shared output folder for Forge Classic/Neo
• Fixed #1466 - crash after moving portable install
• Fixed #1445 - Linux app updates not actually updating - thanks to @NeuralFault!

Supporters

🌟 Visionaries

To our stellar Visionaries: Waterclouds, JungleDragon, bluepopsicle, Bob S, and whudunit! Your generosity keeps this project thriving and gives us the confidence to tackle the big challenges. Thank you for being the foundation that makes it all possible!

🚀 Pioneers

Shoutout to our incredible Pioneer crew for keeping the momentum going! Thank you to: Szir777, Noah M, [USA]TechDude, Thom, SeraphOfSalem, Desert Viber, Adam, Droolguy, ACTUALLY_the_Real_Willem_Dafoe, takyamtom, robek, Ghislain G, Phil R, Tundra Everquill, and a warm welcome to our newest Pioneers: Andrew B, snotty, Miguel A, SinthCore, and Ahmed S!

v2.15.5

Added

• Added new package - Wan2GP
• Added Stable Diffusion WebUI Forge - Neo as a separate package for convenience
• Added Tiled VAE decoding as an Inference addon thanks to @NeuralFault!
• Added togglable --uv argument to the SD.Next launch options

Changed

• Moved the original Stable Diffusion WebUI Forge to the "Legacy" packages tab due to inactivity
• Updated to cu130 torch index for ComfyUI installs with Nvidia GPUs
• Consolidated and fixed AMD GPU architecture detection
• Updated SageAttention installer to latest v2.2.0-windows.post4 version
• Updated torch index for Forge-based UIs to rocm6.4 / cu128 depending on GPU - thanks to @NeuralFault!

Fixed

• Fixed #1450 - Older SD.Next not launching due to forced --uv argument
• Fixed duplicate custom node installations when installing workflows from the Workflow Browser - thanks again to @NeuralFault!

Special Thanks

🌟 Visionaries

To our incredible Visionaries: Waterclouds, JungleDragon, bluepopsicle, Bob S, and whudunit! Your generosity drives this project forward and helps us turn ideas into reality. Thank you for being such a vital part of Stability Matrix!

🚀 Pioneers

A massive thank you to our Pioneer crew! Your support keeps the lights on and the updates flowing. Shoutout to: Szir777, Noah M, [USA]TechDude, Thom, SeraphOfSalem, Desert Viber, Adam, Droolguy, ACTUALLY_the_Real_Willem_Dafoe, takyamtom, robek, Phil R, Tundra Everquill, TheTekknician, and a warm welcome to our new Pioneers, Benjamin M and Ghislain G!

v2.15.4

Changed

• Updated Early Access indicators in the Civitai Details page to be more visible
• Updated error message when attempting to download a website-generation-only model from Civitai
• Updated nunchaku installer to 1.0.2
• Updated Package Import dialog to have Python version selector

Fixed

• Fixed #1435 - SwarmUI not launching due to missing dotnet
• Fixed various install and Inference issues with ComfyUI-Zluda - big thanks to @neural_fault for the PRs!
• Fixed sageattention version getting out of sync after torch updates in ComfyUI
• Potentially fixed issue where uv-managed Python versions would not appear in the version selector

Supporters

🌟 Visionaries

Our heartfelt thanks to the driving force behind our progress, our Visionaries: Waterclouds, JungleDragon, bluepopsicle, Bob S, and whudunit! Your incredible support is the fuel that powers our development, allowing us to tackle bugs and push forward with confidence.

🚀 Pioneers

A huge shoutout to our amazing Pioneers, who keep the momentum going strong! Thank you for being our trusted crew on this journey: Szir777, Noah M, USATechDude, Thom, SeraphOfSalem, Desert Viber, Tundra Everquill, Adam, Droolguy, Philip R., ACTUALLY_the_Real_Willem_Dafoe, takyamtom, and robek!

v2.15.3

Changed

• Updated fallback rocm index for InvokeAI to rocm6.3
• Updated SwarmUI to launch via the launch script for better compatibility

Fixed

• Fixed cuDNN frontend error on ComfyUI-Zluda startup (thanks @neural_fault!)
• Maybe finally actually fixed threading issue with the Python Packages dialog search box for real this time? (may fix #1392)
• Fixed potential install failures when moving duplicate files into shared model folders (may fix #1393)
• Fixed potential threading issues with the Inference image gallery (may fix #1408)
• Fixed #1424 - Civitai account 401 error when connecting accounts, updated for new API changes

Supporters

🌟 Visionaries

Our deepest gratitude to our Visionaries for their foundational support: Waterclouds, JungleDragon, bluepopsicle, Bob S, and whudunit! Your commitment allows us to focus on the essential work of squashing bugs and improving stability, ensuring a rock-solid experience for everyone.

🚀 Pioneers

A huge thank you to our incredible Pioneers for keeping the project on track! Your support is vital for these important refinement updates. Thank you to Szir777, Noah M, USATechDude, Thom, SeraphOfSalem, Desert Viber, Tundra Everquill, Adam, Droolguy, Philip R., ACTUALLY_the_Real_Willem_Dafoe, takyamtom, and a warm welcome to our newest Pioneer, robek!

v2.15.2

Changed

• Updated Avalonia to 11.3.7

Fixed

• Fixed #1409 - incorrect triton version installed during FramePack install on non-Windows systems
• Fixed #1410 - crash when resizing Civitai model details page
• Fixed #1417, #1419 - unable to connect Inference to ComfyUI after updating to latest ComfyUI
• Fixed missing dependencies for ComfyUI API nodes

Supporters

🌟 Visionaries

Huge thanks to our incredible Visionaries: Waterclouds, JungleDragon, bluepopsicle, Bob S, Ibixat, and whudunit! You’re the reason Stability Matrix keeps pushing forward. Your support lights the path and keeps the dream alive.

🚀 Pioneers

Shoutout to our amazing Pioneer crew: Szir777, Tigon, Noah M, USATechDude, Thom, SeraphOfSalem, Desert Viber, Tundra Everquill, Adam, Droolguy, Philip R., ACTUALLY_the_Real_Willem_Dafoe, and takyamtom! You help keep the gears turning and the magic flowing.

v2.15.1

Changed

• Upgraded ComfyUI-Zluda and Stable Diffusion WebUI AMDGPU Forge packages to install and use HIP SDK 6.4.2
• Changed ComfyUI torch index from cu129 back to cu128 for better compatibility with custom nodes
• Updated SageAttention installer to install v2.2.0-windows.post3
• Updated Nunchaku installer to install v1.0.1
• Updated Windows ROCm ComfyUI installs to use recommended environment variables by default

Fixed

• Fixed #1372, #1399 - LiteAsyncException upon starting Stability Matrix v2.15.0
• Fixed #1391 - "Failed to parse" error when upgrading pip packages with extra index url
• Fixed #1401 - "Python was not found and/or failed to install" errors when path contains special characters
• Fixed #1403 - Checkpoint Manager filters not being saved correctly
• Fixed #1411 - SD.Next installs not using correct torch version
• Fixed "cannot access local variable 'job' where it is not associated with a value" error when running jobs in AI Toolkit
• Fixed Civitai browser not always returning at least 30 results when possible on initial search
• Fixed model browser crashing when downloading a file with invalid characters in the name
• Fixed model browser crashing when no author exists for a model

Supporters

🌟 Visionaries

To our guiding stars, the Visionaries! Thank you Waterclouds, JungleDragon, bluepopsicle, Bob S, Ibixat, and whudunit! While this release is focused on fixes and stability, your foundational support is what empowers us to build a reliable and robust platform for everyone.

🚀 Pioneers

A huge round of applause for our fantastic Pioneers! Your steady support helps us smooth out the rough edges and deliver a better experience with every update. Our deepest thanks to: Szir777, Tigon, Noah M, USATechDude, Thom, SeraphOfSalem, Desert Viber, Tundra Everquill, Adam, and Droolguy. We're also thrilled to welcome our newest Pioneers to the crew: Philip R., ACTUALLY_the_Real_Willem_Dafoe, and takyamtom!