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!

v2.15.0

Added

• Added new package - AI Toolkit

• Added new package - FramePack
• Added new package - FramePack Studio

• Added Python Version selector for all new package installs

• Added the ability to rename packages
• Added support for authenticated model downloads in the HuggingFace model browser. Visit Settings → Accounts to add your HuggingFace token.

• Added support for dragging-and-dropping Civitai-generated images into Inference to load metadata
• Added the ability to search by pasting an entire Civitai model URL into the search bar in the Civitai model browser
• Added "Clear Pip Cache" and "Clear uv Cache" commands to the Settings -> Embedded Python section
• Added settings to disable base models from appearing in the Checkpoint Manager and Civitai Model Browser base model selectors
• Added Inference "Favorite Dimensions" quick selector - editable in Settings → Inference, or click the 💾 button inside the dropdown

• Added setting for Inference dimension step change - the value the dimensions increase or decrease by when using the step buttons or scroll wheel in Inference
• Added "Install Nunchaku" option to the ComfyUI Package Commands menu
• Added "Select All" button to the Installed Extensions page
• Added experimental ROCm pytorch install for ComfyUI (non-Zluda) on Windows - requires a compatible AMD GPU
• Added base model type labels (SD1.5, SDXL, Flux, etc.) to Inference model selection boxes

• Added UNET shared folder link for SD.Next
• Added Manual Install button for installing Package extensions that aren't in the indexes
• Added Next and Previous buttons to the Civitai details page to navigate between results
• Added Negative Rejection Steering (NRS) by @Reithan to Inference
• Added Wan 2.2 models to the HuggingFace tab of the model browser
• Added Tiled Encode/Decode options to FaceDetailer in Inference
• Added Ukrainian translation thanks to @r0ddty!
• Added Czech translation thanks to @PekarT!

Changed

🌟 Civitai Model Details: A Grand Reimagining! 🌟

• No more peering through a tiny window! Introducing a massive overhaul of the Civitai Model Details page, transforming it from a cramped dialog into a spacious, feature-rich hub for all your model exploration needs.
• We've listened to your howls for more, and now you can dive deep into every aspect of your favorite models with unprecedented clarity and control:
  • Expansive View: The new full-page layout means all essential information, descriptions, and previews are laid out beautifully, banishing the old, restrictive dialog forever.
  • Rich Details at a Glance: Author, base model, last updated, SHA hashes, file name overrides/patterns – everything you need, perfectly organized and always accessible.
  • Overhauled Image Viewer: Enjoy a sleek, modern image viewer that includes Civitai metadata and supports zooming, panning, and full-screen viewing. No more squinting at tiny thumbnails!
  • Integrated Inference Options: For supported models, adjust sampler, scheduler, steps, CFG Scale, width, and height directly from the details page, streamlining your workflow like never before!

• Updated all Python version management, virtual environment creation, and pip installs to use uv for improved reliability, compatibility, and speed

• You can now select release versions when installing ComfyUI
• You can no longer select branches when installing InvokeAI
• Updated InvokeAI install to use the intended install method (resolves #1329)
• Updated ComfyUI installs for AMD users on Linux to use the latest rocm6.3 torch index
• Updated ComfyUI-Zluda installs to use the newer install-n method (fixes #1347)
• Removed disclaimer from reForge since the author is now active again
• Updated git operations to better avoid conflicts
• Updated Japanese translation
• Civitai model browser image loading now uses dynamic resizing for better performance and a smoother scrolling experience
• Undo ComfyUI process tracking changes for now due to causing more issues than it solved
• Updated GPU parsing fallback on Linux systems to use the method provided by @irql-notlessorequal
• New installs of ComfyUI, SD.Next, and InvokeAI will now use Python 3.12, unless otherwise specified in the Advanced Options during installation
• New installs of all other packages will now use Python 3.10.18, unless otherwise specified in the Advanced Options during installation
• Updated ComfyUI installs for AMD users on Linux to use the latest rocm6.4 torch index
• Updated package delete confirmation dialog

Fixed

• Fixed an error when packages and other processes exit before process tracking on windows can initialize
• Fixed "none" appearing in wildcard field when using Face Detailer in Inference
• Fixed #1254 - Unable to scroll samplers in Inference
• Fixed #1294 - Improper sorting of output folders in Output Browser
• Fixed #1300 - Git errors when installing Extension Packs
• Fixed #1317 - Inference missing GGUF text encoders
• Fixed #1324 - Window height slightly increasing every launch
• Fixed #1357 - Case insensitivity causing duplicate key exceptions on non-Windows systems
• Fixed #1360 - A1111 install not using correct torch for 5000-series GPUs
• Fixed #1361 - numpy and other Forge startup
• Fixed #1365 - Output folder list not updating when Refresh button clicked

Supporters

🌟 Visionaries

To our incredible Visionaries, the architects of our ambition: Your profound support is the powerhouse behind this massive v2.15.0 release. You don't just light the path; you fuel the entire journey, allowing us to build bigger, move faster, and turn bold ideas into reality. Our deepest gratitude to: Waterclouds, Corey T, bluepopsicle, Bob S, Ibixat, whudunit, and TheTekknician! We are immensely grateful for your trust and partnership in shaping the future of Stability Matrix. Thank you for everything!

🚀 Pioneers

A heartfelt salute to our trailblazing Pioneers! Your consistent support helps us navigate the development landscape, ensuring we stay on the right track and can explore new frontiers. A huge thanks to: tankfox, Mr. Unknown, Szir777, Tigon, Noah M, USATechDude, Thom, SeraphOfSalem, and a special welcome to our newest Pioneers - Desert Viber, Tundra Everquill, Adam, and Droolguy! Thank you for being the vanguard of our community!

v2.14.3

Added

• Added the ability to search by pasting an entire Civitai model URL into the search bar in the Civitai model browser

Changed

• The main sidebar now remembers whether it was collapsed or expanded between restarts.
• Inference is now able to load image metadata from Civitai generated images via drag & drop
• Updated process tracking for ComfyUI to help mitigate restart issues when using Comfy Manager
• Updated pre-selected download locations for certain model types in the Civitai model browser
• Updated nodejs to v20.19.3 to support newer InvokeAI versions

Fixed

• Fixed missing .NET 8 dependency for SwarmUI installs in certain cases
• Fixed ComfyUI-Zluda not being recognized as a valid Comfy install for the workflow browser
• Fixed #1291 - Certain GPUs not being detected on Linux
• Fixed #1284 - Output browser not ignoring InvokeAI thumbnails folders
• Fixed #1301 - Error when installing kohya_ss
• Fixed #1305 - FluxGym installing incorrect packages for Blackwell GPUs
• Fixed #1316 - Errors when installing Triton & SageAttention
• Fixed "directory is not empty" error when updating packages with symlinks
• Fixed missing base model types in the Checkpoint Manager & Civitai Model Browser

Supporters

🌟 Visionaries

Big heartfelt thanks to our stellar Visionary-tier Patrons: Waterclouds, Corey T, bluepopsicle, Bob S, Ibixat, and whudunit! 🌟 Your extraordinary generosity continues to fuel Stability Matrix’s journey toward innovation and excellence. We appreciate you immensely!

🚀 Pioneers

Massive thanks to our fantastic Pioneer-tier Patrons: tankfox, Mr. Unknown, Szir777, Tigon, Noah M, USATechDude, Thom, and SeraphOfSalem! Your unwavering support keeps our community thriving and inspires us to push even further. You’re all awesome!

v2.14.2

Changed

• Changed Nvidia GPU detection to use compute capability level instead of the GPU name for certain feature gates / torch indexes

Fixed

• Fixed #1266 - crash when moving or deleting Lora models in the Checkpoint Manager
• Fixed #1268 - wrong torch index used for Nvidia 1000-series GPUs and older
• Fixed #1269, #1257, #1234 - "no such file or directory" errors when updating certain packages after folder migration
• Fixed #1274, #1276 - incorrect torch installed when updating to InvokeAI v5.12+
• Fixed missing shared folder links for SwarmUI's diffusion_models and clip folders

Supporters

🌟 Visionaries

Our deepest gratitude to the brilliant Visionary-tier Patrons: Waterclouds, bluepopsicle, Bob S, Ibixat, and Corey T! Your incredible backing is instrumental in shaping the future of Stability Matrix and empowering us to deliver cutting-edge features. Thank you for believing in our vision! 🙏

🚀 Pioneers

A huge shout-out to our fantastic Pioneer-tier Patrons: Mr. Unknown, tankfox, Szir777, Noah M, USATechDude, Thom, TheTekknician, and SeraphOfSalem! Your consistent support and active engagement are vital to our community's growth and our ongoing development efforts. You truly make a difference! ✨

v2.14.1

Changed

• Updated Inference Extra Networks (Lora / LyCORIS) base model filtering to consider SDXL variations (e.g., Noob AI / Illustrious) as compatible, unrecognized models or models with no base model will be considered compatible.
• Changed hotkey for Inference prompt weight adjustment to (⌘+Up/⌘+Down) on macOS
• Improved style consistency of Inference Prompt action buttons on top right
• (Internal) Improved log console formatting & colorization for development

Fixed

• Fixed Inference hotkey weight adjustment multi-line behavior, now works as expected like the first line.
• Fixed updates to versions with commit hash version parts not being recognized when the current version has no commit hash part.
• Fixed Inference Extra Networks card not updating with newly added model files.
• Fixed incorrect ROCmLibs being installed for RX 6800/6800XT users of Comfy-Zluda or AMDGPU-Forge
• Fixed missing text when missing localized versions for Italian and Chinese languages
• Fixed Python Packages dialog errors and potentially other issues due to concurrent OnLoaded events

Supporters

🌟 Visionaries

Big cheers to our incredible Visionary-tier Patrons: bluepopsicle, Bob S, Ibixat, Waterclouds, and Corey T! 🚀 Your amazing support lets us dream bigger and reach further every single month. Thanks for being the driving force behind Stability Matrix - we genuinely couldn't do it without you!

🚀 Pioneers

Huge thanks to our fantastic Pioneer-tier Patrons: tankfox, Mr. Unknown, Szir777, Tigon, and Noah M! Special shoutout and welcome back to TheTekknician, and a warm welcome aboard to our newest Pioneers: USATechDude, SeraphOfSalem, and Thom! ✨ Your continued support keeps our community vibrant and pushes us to keep creating. You all rock!