Preview Release 2

^b2000

Download SwiftDiffusion.app

Tip

Instructions

Drag and drop into your /Applications folder.

---

Important

Updating from Preview Release 1?

Due to the nature of this project being in early development, things are constantly changing to be more efficient–including the local database setup. As such, until the data structure has been finalized, this means that old database schemas may become incompatible with newer database configurations. This is only temporary.

But worry not: You'll still be able to access your existing database by re-downloading Public Release 1. You can also transfer prompts manually, from one version of the app to another, using the Copy/Paste feature in the Prompt menu bar.

---

Note

Regarding CoreML

SwiftDiffusion CoreML^BETA is planned to hit the preview release cycle soon. Hold tight!

---

What's New in Preview Release 2?

• Lots and lots and lots of stability improvements
  • Seriously, lots
• Improved main thread priority processes
• Reduced main thread processes to essential tasks
• Improved background, asynchronous logic
  • Most processes should no longer stall on main
  • Still refactoring @MainActor logic for essential model interfaces
• Fix rendering, lag issues on window active
• Fix stutter issues with PyTorch processes
• Fix invalid sampling method key in JSON payload
• Improved local database model management
  • Non-optionals, non-deletables: rootFolder, workspaceFolder
  • .parent relationship attributes
  • @relationship(deleteRule: .cascade)
• Improved SidebarItem management
  • Centralize SidebarModel logic
  • Fix item deletion logic in context
  • Additional handling logic for deletion (recursive delete ImageInfo, PromptModel, cache, etc.)
• Improved Folder management
  • Centralize SidebarModel logic
  • Fix folder deletion logic in context
  • Recursive handling for embedded folders, items (queued for deletion)
• Reworked Parse Generation Data logic from pasteboard
  • Asynchronous checks for valid tags
  • Parsing creates a new workspace item now instead of interfacing with PromptView directly
• Improved copy, move logic
• Add persisting "New Prompt", "New Folder" below each section
• Fix Checkpoint Manager issues
  • Fix nil sheet view causing unclosable window
  • Fix deletion, presentation direct from model items
  • Alphabetical ordering of checkpoints
    ...

Read the full changelog here.

Technical Changelog: v0.1-beta.0...v0.2-beta.0

---

What can I look forward to in the near future?

CoreML implementation

• Integrate into existing PyTorch interface
• PyTorch-to-CoreML model checkpoint converter
• Prepare modular implementation for rumored AI
• Expand on existing features, as seen with PyTorch

Building on PyTorch

• Highres, Refinement
• ControlNet (IPAdaptor, etc.)
• Img2Img, Inpainting
• Plugin support
• Every Automatic feature
• ComfyUI, Forge

General, QoL

• Multi-prompt queue
• In-depth model checkpoints management
• Provide option to auto-download missing LoRAs, VAEs, etc. included in prompt
• iOS & iPadOS views for queuing and organizing prompts on mobile via local network

Much, much more.

Preview Release 1

^b1000

Download SwiftDiffusion.app

Instructions

Drag and drop into your /Applications folder.

---

Welcome to the first public preview release of SwiftDiffusion! Enjoy!

Updated for Automatic v1.8.0

For the time being, you'll need an existing installation of the Automatic webui script. I hope to improve the onboarding process in the future. In the meantime, try to break things so that we can create a better product!

Furthermore, please share your workflows and how you interact with stable diffusion on your day-to-day so that I can create a better user experience.

Finally, keep in mind that this is a preview release. All of the foundation has been laid to implement every feature currently present in Automatic. Features will be added as stability is tested. However, these initial few releases will be kept simple for the time being to ensure that there is a stable enough base to build upon.

---

What can I look forward to in the near future?

CoreML implementation

• Integrate into existing PyTorch interface
• PyTorch-to-CoreML model checkpoint converter
• Prepare modular implementation for rumored AI
• Expand on existing features, as seen with PyTorch

Building on PyTorch

• Highres, Refinement
• ControlNet (IPAdaptor, etc.)
• Img2Img, Inpainting
• Plugin support
• Every Automatic feature
• ComfyUI, Forge

General, QoL

• Multi-prompt queue
• In-depth model checkpoints management
• Provide option to auto-download missing LoRAs, VAEs, etc. included in prompt
• iOS & iPadOS views for queuing and organizing prompts on mobile via local network

Much, much more.