Well! Hi there, it's been a minute. Welcome to 0.6.2 Beta -- this one got a rather long time before I finalized the release, as I wanted the "Simple" tab ready to go first. So as a side effect... this update has more new features than any other release thus far, and more bugfixes/QoL improvements/etc than I can even be bothered to try to count.

Major Updates

• New Themes: "Modern Dark" and "Modern Light", created by Eric and pythongosssss -- https://github.com/Stability-AI/StableSwarmUI/discussions/11#discussioncomment-9153506
• new "SwarmInput" comfy nodes to allow fully specifying parameter details when building comfy workflows for usage with swarm
• added new "Simple" tab, which lets you build custom workflows and provide an easy direct-linkable interface that's safe, clean, and easy for even beginners / non-technical people to use your workflows with
• Added automatic installer buttons for controlnet and ip-adapter
• added custom word autocompletions setting (eg for anime booru tags) -- https://github.com/Stability-AI/StableSwarmUI/blob/master/docs/Autocompletions.md
• added prompt alternating steps syntax a photo of a <alternate:cat,dog>, and a prompt-editing-timestep syntax a photo of a <fromto[0.2]:green, blue> cat -- https://github.com/Stability-AI/StableSwarmUI/discussions/11#discussioncomment-9206140

Minor Updates

• made a "Sampling" group to hold the Sampler & Scheduler, finally
• added hacky placeholder "ControlNet Two" and "Three" groups for if multiple controlnets are needed
• a variety of minor visual aesthetic tune-ups while working on the "Modern" themes
• the negative prompt box is now by default in the center under the regular prompt box
• added an 'auto swap to images' checkbox (gear icon at top-right, where the 'auto swap to previews' checkbox is)
• added support for CosXL
• added support for CosXL Edit (ip2p) - just use the model as normal, and give an Init Image. then write a prompt like "give him a mustache" to instruct the edit model on what to change
• wrote a doc guide about how to make extensions for swarm: https://github.com/Stability-AI/StableSwarmUI/blob/master/docs/Making%20Extensions.md
• added "always refresh on load" setting, for performance benefit reasons. Disable it if your swarm instance is lagging from loading too many models or much data out of comfy or similar.
• added "backend data cache" setting, also for performance. Enable it if your comfy instance takes a long time to load the UI, or you get timeouts from object_info loads.
• when you edit model paths in server settings, it now (finally!) reloads the model handler immediately to apply your changes without restarting the program
• cleaned up model name displays so it's not spamming ".safetensors" on the end of every single one
• added automatic log save-to-file as an option under Server Configuration
• pushed more work towards making a functional multi-user support system
• added a ModelSpecificEnhancements under Swarm Internal that you can uncheck to disable SDXL rescond smartness and possibly other enhancements like it in the future
• added advanced selection for Mask Behavior inside init image, defaulting to Differential now with the ability to swap back to Simple Latent. This makes masked image gen (including eg <segment: usage) a bit cleaner/better
• added align_your_steps scheduler, optimal for 10 steps but can do less, useful especially for coherent SVD outputs without taking a year to process
• added Mask Shrink Grow parameter, for when using image editor / masks, to crop the image to just the mask, then grow the mask by a few pixels, and generate based on that. This is useful for refining small details.
• model descriptions can now use (heavily restricted whitelist-only) html tags to add a bit of formatting
• added a check-for-updates when swarm is starting. No telemetry, no downloading anything, just grabs a json off github to compare version number with. If you don't want this, you can disable it in server settings.
• your selected tab is now in the URL, so you can bookmark specific tabs to get back to them with fewer clicks
• number inputs now let you scrollwheel to quickchange (firefox-like behavior on all browsers), or click-and-drag
• the image full view (when you doubleclick an image) can now be zoomed with mousewheel and panned with click-n-drag, to let you investigate details easily
• published initial API docs https://github.com/Stability-AI/StableSwarmUI/blob/master/docs/API.md (note some routes don't have descriptions written yet, and OpenAPI stuff is planned)
• as usual, many small improvements, bug fixes, better error checking, etc.

Beta!

Swarm is now in Beta status! After months of "Alpha" status, I'm ready to say Swarm is one of the best Stable Diffusion UIs out there, and for many use cases simply the best.

• Brand new to Stable Diffusion? Swarm is crazy easy to install and set up and will literally guide you through how to get started in-UI.
• Inexperienced? There's a ? on every param to tell you how each feature works, and most features you might care about are available out of the box. There's tab completions for prompt syntax. Everything is ready to help you.
• Pro and want a friendly UI? Swarm's main Generate tab is one of the friendliest out there.
• Pro and want to maximize your comfy noodle hackery? Swarm wraps comfy in a tab and provides power features such as a workflow browser, the ability to generate reference workflows for you, the ability to reformat your favorite workflow into a friendly interface, large scale gridding, multi-GPU support, ...
• Large scale commercial user? Swarm provides a great frontend for you to start building with and debug with, and a powerful backend + splitting engine + etc., and a friendly license.

Major Updates

• We've moved to the MIT License, meaning Swarm itself is now free to use for all users, even commercially. Note that models have their own licenses still.
• Added multi-language support! You can now select languages from the list menu at the top of the UI - feature announcement: https://github.com/Stability-AI/StableSwarmUI/discussions/11#discussioncomment-8131492
• new theme added in a PR by QuantumSoul
  • and the existing themes now look a little better and respond well to a wider parameters view, thanks to the same PR
• Stable Cascade supported natively https://github.com/Stability-AI/StableSwarmUI/discussions/11#discussioncomment-8521943
• added a Comfy Workflow browser to the top of the comfy tab, now you can easily browse your saved workflows - https://github.com/Stability-AI/StableSwarmUI/discussions/11#discussioncomment-8736304
• Updated core to .NET 8, removing the dependency on .NET 7

Minor Updates

• video by default now has smarter parameters that result in better quality videos with SVD
• added Refiner Steps param
• refiner save-before-refining is now a general "Save Intermediate Images" that can also save before segmentation
• added automatic recognition for "Matrix" app formatted model metadata files
• gridgenerator now has convenient autoreplacables in web-page output folder name
• experimental AMD GPU auto-support in installer. Doesn't work great.
• added "Video Boomerang" advanced option checkbox
• added video resolution parameter
• added native support for frame interpolation from https://github.com/Fannovel16/ComfyUI-Frame-Interpolation in video params
• added user setting to exclude params from the "Reuse Parameters" button
• added swarm internal 'image format' parameter override for convenience especially with api usage
• added a close button to the image editor (most requested feature of 2024 lol)
• the image editor has a working selector tool now, which also can be used to automatically create a regional prompt with a simple button press
• added support for GLIGEN - only for SDv1 models, under Advanced parameters -> Regional Prompting, you can select a model as long as you have one installed per https://comfyanonymous.github.io/ComfyUI_examples/gligen/
• added image history sort-by (Name vs Date and a Reverse option)
• new parameter: ShiftedLatentAverageInit under advanced sampling - empty latents are normally initialized as zeroes, but that's not the correct average value for VAEs. Check this box to use actual channel-average values (based on FFHQ dataset as calculated by Birchlabs https://huggingface.co/datasets/Birchlabs/sdxl-latents-ffhq)
• added <break> syntax to separate conditioning manually to multiple parts
• improved the install and launch scripts for all platforms (no longer rebuilds the C# code every launch, now only when you update or launch with the dev mode powershell script)

--

ps yes of course when SD3 comes out Swarm will support it right away.

Plans

Swarm is gearing up for a transition out of Alpha status into Beta. Final work to do before it's ready for "Beta" includes documentation, translations, and finishing up the initial image editor tool. The Beta will most likely also be relicensed as the fully FOSS MIT License.

Swarm is already at a point where it's one of the best possible UIs you can have for Stable Diffusion, and there's very few reasons to not use Swarm. Reaching Beta will mark the point where the reasons to use anything other than Swarm are almost nil. I'm incredibly proud of the progress this project has made over the past months in Alpha.

Major Updates

• A lot Grid Generator enhancements - https://github.com/Stability-AI/StableSwarmUI/discussions/11#discussioncomment-7862586
  • The GridGenerator tool can now generate "Just Images" or an "Image Grid" as an alternative to the webpage viewer thingy
  • The GridGenerator tool can now save/load configuration state
• Added Self-Attention-Guidance parameter options, under Advanced Sampling
• The 'Star' button on images now actually does something (saves a copy to the "Starred" folder). There's a usersetting for whether it'll use subfolders or not. Also highlights the starred images in the UI.
• DotNET 8 was released recently, and swarm will update to it in the near future. I recommend you install DotNET 8 in advance to make sure the update is painless when it comes https://dotnet.microsoft.com/en-us/download/dotnet/8.0
• made the installer a lot nicer (both the install UI experience, and a proper Linux install script) - https://github.com/Stability-AI/StableSwarmUI/discussions/11#discussioncomment-7936481
• new parameter under Init Image group: Unsampler Prompt, use to unsample your input image before resampling. This replaces the random noise with noise based on the input image + a prompt, allowing for more tightly controlled image editing
• added a parameter filter at the top to more easily find obscure params

Minor Updates

• added a user setting to allow creating grids of images when you generate a bunch at once (eg a 2x2 grid when you generate 4 at once) (default disabled)
• added a user setting for image history length (default max 1000 images)
• image history now caches small preview thumbnails by default (you can disable this in user settings)
• added advanced swarm internal param no seed increment if for example you want to muck with wildcards and not have seed change, or whatever
• Grid Generator now has on-page docs and a link to the readme about it
• Grid Generator now highlights invalid values
• Grid Generator now supports skipping parameters with a SKIP: prefix
• Image Edit Batcher now has image resolution controls
• dropdowns now have fancy search text input
• backends list now has "Restart" and "View Logs" buttons on each backend for convenience
• Added No Previews advanced param under Swarm Internal. Implicitly enabled by default for grid generations. May make some generations slightly faster.
• Added Refiner Save Before Refine advanced param (under Refiner)
• video generation now has an option for different preview types, including live previews of the animated video while it generates
• there's now memory freeing controls - API route FreeBackendMemory, buttons on Server Info tab, and server settings for autoclear after specified time
• there's now a resource usage monitor on the server info tab
• as always, a bunch of other fixes n small things

Major Updates

• New param: InitImageResetToNorm, lets you overdrive InitImaageCreativity(denoising) above 1.0 by mixing the init with latent emptiness
• Stable Video Diffusion (SVD) is now supported with a "Video" param group, for img2vid and txt2img2vid, see https://github.com/Stability-AI/StableSwarmUI/discussions/11#discussioncomment-7668336 for details
• added a Logs tab under Server, so you can look through the raw server logs if needed entirely in UI. This also lets you view the Debug logs without having to enable them in the main terminal.
• New LoRA Extractor utility to get shrink thicc checkpoint models into smol loras https://github.com/Stability-AI/StableSwarmUI/discussions/11#discussioncomment-7772593
• Prompt input areas now have tab completion for prompt syntax (eg wildcards, embeds, etc)! https://github.com/Stability-AI/StableSwarmUI/discussions/11#discussioncomment-7775593
• added multi-random/repeat prompt syntax magic - see https://github.com/Stability-AI/StableSwarmUI/issues/196#issuecomment-1848391633 for details
• have an in-progress Image Editor tool. It's not quite ready for primetime, but you can poke at it if you're curious. It's a full multilayered canvas-based editing suite.

Minor Updates

• added restart all backends quick button
• added support for Cubiq's IPAdapter repo
• new parameter: RemoveBackground under advanced sampling, to just quickly remove the background from your images
• new prompt tag option: <clear:x> works like the <segment: node, but removes the content from the image - just like RemoveBackground but targetable with ClipSeg.
• new parameters: ControlNet Start and End, to control which steps a controlnet applies to (only visible if Advanced params is checked)
• TI Embeddings now have a metadata key that lets you mark them as Negative, so that when you click them they will be automatically added to the negative prompt rather than the positive
• added a user-setting to select the default VAE for SDXL and for SDv1 separately. eg for SDXL you might want the fp16-fix model.
• added Refiner HyperTile param (a technique that makes sampling a bit faster on large images)
• When you right click Generate and click Generate Previews, it will now try to use a preset named Preview, and can even help you generate one if you have an sdxl_lcm_lora available already
• the "Generate" button in the center area now has a more-obvious clickable menu button for additional options
• there's now a "Show Prompt Tokenization" in that generate-options button to quickly view the clip tokenization details for it (ie via the utilities tab)
• oh and as usual a bunch of random small fixes and wotnot

Major Updates

• added support for animated GIF outputs, from eg AnimateDiff or ComfyVideoHelperSuite nodes. Swarm now recognizes and handles these properly throughout, including the grid generator and everything.
  • Video files (mp4/webm) are now partially supported as well.
• Added Docker support, see readme for info
• Added a bulk pickle-to-safetensors conversion utility, see here for detail
• added support for remote Swarm backends having their own model lists recognized by the local Swarm instance, including automatically filtering requests to the correct backend that has the desired models, and support for model/lora/embedding/controlnet all being recognized
• you can now drag an image into the image area and it will be loaded in and metadata scanned - this includes images created by swarm, or even images created by other tools (eg auto webui or fooocus)
• ComfyUI Workflow Editor tab can now use multiple backends simultaneously! At the top left under MultiGPU selector choose All, then create a workflow where the output nodes (PreviewImage/SaveImage) have different colors (use 1 unique color per backend). Each output's sub-graph will be split to a different comfy backend - see video here https://github.com/Stability-AI/StableSwarmUI/discussions/11#discussioncomment-7555872
• added wildcards system - for details see https://github.com/Stability-AI/StableSwarmUI/discussions/11#discussioncomment-7568190

Minor Updates

• when using a custom Comfy workflow in the main generate tab, you can add a new SwarmLoraLoader node to let the main UI lora selector work, see here for details
• added VAE Tile Size param under Advanced Sampling for if you need VAE tiling
• added a mass model metadata reset utility
• added automatic recognition of Embed model architectures
• grid gen axis tab completion now offers an 'Add All' to quick-add a range of matches
• grid gen now automatically reorders axes internally to maximize speed (eg changing model lastmost since model changes are slow) (and has a checkbox to turn this off if you don't want it to)
• added used_embeddings to metadata when you use TI embeddings
• added a button to manually clear the batch view (under the gear icon, at the top-left of the batch view area). Also the autoclear toggle now defaults to off.
• there's now a button in the Models listing (under the hamburger menu) to Set As Refiner quickly
• there's now a user setting to change the hint popovers (the ? clickies) between buttons, hoverables, or none (hidden)
• image metadata is now saved in a format that's recognized by other tools (PNG Text Values for png, or UserComment EXIF for jpg)
• added CLIP Stop At Layer under Advanced Sampling for if you want that for old SDv1 NAI style model
• added Personal Note parameter, under advanced->swarm internal, if you want to handwrite a note into the metadata
• added Open In Folder button on the hamburger menu for images
• as always, a lot of assorted super minor things not worth listing

Major Updates

• added segmentation-detailing - see https://github.com/Stability-AI/StableSwarmUI/discussions/11#discussioncomment-7236821 for details, but in short: you can do <segment:face> a beautiful face on the end of a prompt to trigger automatic inpainting of segments matched by text
• ReVision can now use IP-Adapter if you have that node in your comfy
• added "Seamless TIleable" checkbox under "advanced sampling" so you can make textures n stuff
• clicking embeddings now actually works and puts it in the prompt and all

Minor Updates

• large amounts of internal stuff n wotnot as always, including some API improvements and load time speedups
• added MaskImage parameter - works as a jank way to inpaint for now, or as API inpainting, will be used to build UI around later
• added new <object:...> syntax to complement <region:...> except object emits additional generation calls in the middle to more aggressively form the wanted image
• image metadata now shifts to the side when doing so allows the image to be bigger
• auto-clear-batch is now under a lil Gear icon
• there's now a toggler to toggle whether to instant-swap to previews or not
• added advanced parameters to modify sigma min/max and rho
• added FreeU Apply To param to control whether freeU applies to base, refiner, or both
• added ReVision Zero Prompt param to force ReVision to zero the prompts given to the base (and leave unzeroed for refiner)

Major Updates

• added prompt syntax - <embedding:filename>, <lora:filename:1>, <random:a, b, c>, <preset:presetname> - see https://github.com/Stability-AI/StableSwarmUI/discussions/11#discussioncomment-7057501 for detail
• Variation Seed is now available for comfyui backend
• added initial experimental regional prompting support (ComfyUI only), works with an initial dirty/simple syntax in the prompt like a photo of a catdog <region:0,0,1,0.5> a photo of a cat <region:0,0.5,1,0.5> a photo of a dog (proper UI for it and more advanced tooling TBA)

Minor Updates

• the usual pile of bugfixes and minor QOL improvements and etc
• made bulk generations fasterer through rapid prequeueing
• adjusted default image save path and reversed the image history sort, so that together your most recent images end up at the top (the settings change won't self-update a pre-existing server settings file - 'reset to default' UI is TBA, but for now you can check The Settings source to reference default values)
• there's now a button to preview ControlNet preprocessor outputs, in the controlnet param group
• you can now use Control+Up/Down arrow when you've selected prompt text to adjust prompt weighting
• added DynamicThresholding support for self-start ComfyUI, or any ComfyUI-API-By-URL that has the DynThresh node
• Added custom comfy nodes (automatically enabled for self-start backends, you'll have to transfer manually for API URL backends - see ComfyUI Docs to do that) that make image loading & saving much cleaner and faster
• Added "FreeU" support (https://arxiv.org/abs/2309.11497, Comfy backend only) as an option under "Advanced"
• The Comfy Workflow Editor tab now has a QuickLoad dropdown so you can more quickly load workflows you use often
• Refiner now has StepSwapNoisy method if you want to use the official-sample-workflow's way of handling noise, which is a bit better for standard refiner (SDXL Base + SDXL Refiner, No upscale) but much worse for anything else.
• The default Refiner method has been changed to PostApply, as that always works, whereas StepSwap only works for some cases. (You can always change your personal defaults at will by just creating a preset and naming it Default!)
• added the rescond tweaks used by Fooocus to get slightly better image quality
• Previews of large-batches can now load concurrently on a comfyui backend, if you use the advanced "batch size" param. This might negatively impact performance when using TAESD.
• Added a checkbox to use AITemplate in comfy if you have that extension - gens 2x faster, but has compatibility issues

Major Updates

• added an "Image Edit Batch" tool next to the grid gen, for if you want to batch edits to a folderload of images
• added a "Utilities" tab at the top that has a CLIP Tokenizer for if you want to see what CLIP's doing with your text
• Swarm can use remote instances of Swarm as its backend (ie it's now easier to set up Swarm across LAN/network)
• LoRA/etc. browsers now recognize if they're the correct model type (SDv1 vs SDXL) - if they don't match, they fade out and get sorted to the bottom, so SDv1 loras aren't in the way while you're generating SDXL images (but they're still there with a fade and a message so you don't get confused where they went)

Minor Updates

• Colab banned webuis :(
• a bunch of small bug fixes and QOL as always. Like a lot. I can't be bothered listing them all, but you can look at the git commits to see em if you want. So many.
• image metadata now holds the Swarm version and current date (alongside params used and gen time)
• when generating images with a comfy backend there are now lil progress bars on the side. A green one shows the main generation (KSampler) progress, and a subtler purple one shows the overall comfy progress (ie how far along the node graph it is)
• you can now customize the internal batch size if you want, as a parameter under advanced - see notes on https://github.com/Stability-AI/StableSwarmUI/issues/84 about that. Most users prefer the "Images" count at the top of core params
• added a toggle for auto-clearing batches
• custom comfy workflows now show Primitive inputs in the main param section, and auto-generated other options under "Advanced", to be less cluttered of an interface
• added an option for comfy api by url to go idle, meaning its connection can go on and off automatically based on whether the remote server is available. Very useful for external servers that you might only sometimes have running
• there's now a CLIP Token counter on prompt boxes so you can see when you're hitting the 75 token barrier
• added a lil guide doc about how to use swarm over LAN or internet https://github.com/Stability-AI/StableSwarmUI/blob/master/docs/Advanced%20Usage.md#accessing-stableswarmui-from-other-devices
• improved rendering of model cards
• made it way more obvious how to close the image fullscreen popup view

0.5.4 Alpha

I haven't written release notes thus far because there's just so much but I'm going to try for this one.

The major updates here and other features are announced as they happen in the Feature Announcement Thread, I highly recommend following that thread!

Major Updates

This is the first release containing the in-progress UI rework. (There's more to come on that still). Things look differenty now! Hopefully you like it better!

This also contains ReVision support. When using SDXL, drag images into the prompt box to have them parsed as part of the prompt. It's pretty neat!

Minor Updates

This release contains a significant amount of minor UI and QOL improvements, and a few fun new features

• You can now set custom titles on backends
• You can now toggle backends on/off (suggested by AIMusicExperiment)
• presets work a bit more cleanly on the inside now, and can do fancier things like have loras in them
• you can now edit all your parameter settings (including you can make the old prompt param visible again this way if you want)
• errors are now tracked better all over the place
• fixes to edge-cases where comfy did not launch properly for a couple users. I don't want to talk about how annoying this bug was to track down. It was an adventure. I cried a few times. But we did it. Thank you WyzBorrero for helping me test way more than any sane person should.
• the image batch view now puts new images at the top, not the bottom, so you always got the latest visible
• added "alt resolution multiplier" to make it easy to do things like grids of different sized images or whatever if you want to be weird with res
• the Grid Generator now has tab completion suggestions
• a whole lot of improvements to the browser UIs (images/presets/models/etc), including you can select new formats like "Small Cards" or "Giant Thumbnails" or etc now
• made the colab more reliable. Thank you dotcsv, rocketpal, nyaneenee, and yoinked for reports and testing related to colab issues.
• model folders now self-generate as a bunch of people were confused where to put loras and all
• you can now use arrow keys to quickly jump between different images in the current generation batch, including both on the generate page and when you click the image to fullscreen view it (suggested by pookyjuice)
• improved persistence (eg of loras) across page reloads
• added a replacement for the icon, it's now an SDXL-generated purple robot bee. It's kinda cute. I need to ask an actual graphic designer to replace it properly lol
• You can now right-click generate to "Generate Forever", and right-click interrupt to "Interrupt All Sessions" (ie kill old queues that are left from tabs you already closed)
• changed timeout tracking to allow you to queue a thing and let it complete days later if you really want giant terrifying queues to run.
• changed default param limits to let you put them ridiculously high if you want. Queue ten thousand images in one go if you want, who am I to stop you?
• custom comfy workflows were made extra smooth

That. is not. A complete list. That's just the highlights. There's a lot more. Do you see why I normally don't write release notes? There's a lot.

Initial public Alpha of StableSwarmUI.