Skip to content

v5.2.0
Compare
Choose a tag to compare
@psychedelicious psychedelicious released this 17 Oct 11:41
· 498 commits to main since this release
v5.2.0
358dbdb

This release includes a number of enhancements and fixes. Standouts include:

  • Support for FLUX ControlNets.
  • Improved UX for common img2img flows in Canvas.
  • Boards may be sorted by name or date.
  • Starter model bundles in Model Manager.
  • Support for bulk image uploads.

** 🚨 Important Update Note: **

During update, you'll now be selecting between different configurations of the Invoke environment that are optimized and dependent on your hardware (specifically, torch-sdp vs. xformers) -- If you have previously used xformers, and update to the version for 3xxx and 4xxx NVidia cards, you'll often experience an error after the update.

You can uninstall xformers before running the update by -

  • Running the Invoke batch script from your existing installation
  • Selecting the "developer console" option
  • Typing pip uninstall xformers

We've also added a fix for Apple Silicon users for the mushy noise issue and updated the installer to skip xformers for GPUs where it causes a performance hit.

Be sure to review the v5 release notes if you haven't already upgraded to v5.

FLUX ControlNets

We now support both XLabs and InstantX ControlNets for FLUX. We've found the Union Pro model substantially outperforms the other models and added it to the starter models. Other models work, but outputs are not as good.

You can use FLUX ControlNets in both Workflows and the Linear UI. We will be adding a union mode control to the Linear UI in a a future release. You can select the union mode in Workflows today.

Canvas img2img Flow

We've made a number of changes to support the common img2img flow in the Canvas:

  • Transform now supports 3 modes:
    • fill (old behaviour): The layer is stretched to fit the generation bbox exactly. Its aspect ratio is not maintained.
    • contain (new default): The layer is stretched to fit the generation bbox, retaining its aspect ratio.
    • cover: The layer is scaled up so that its smallest dimension fits the bbox exactly, retaining its aspect ratio.
  • Add a layer context menu item to fit the layer to the bbox using the contain mode.
  • Update theNew Canvas from Image image context menu item to streamline the img2img flow. It now resizes the bbox to match the image's aspect ratio, respecting the currently-selected models' optimal size. The image will fit exactly in the box. You can click this and then immediately Invoke to do img2img.

Installer Updates

The performance of torch-sdp attention is substantially faster than xformers on 30xx and 40xx series GPUs. We've made two changes to ensure you generate with the best settings:

  • Add an installer option for 30xx & 40xx series GPUs, which does not install xformers.
  • When the attention type is set to auto (the default), and you do have xformers installed, we choose the best option of torch-sdp and xformers, based on your GPU.

Apple Silicon Fix

The mushy noise issue on Apple Silicon is related to sliced attention. We've temporarily forced all MPS devices to use torch-sdp. Memory usage is a bit higher, but you won't get mushy noise (unless, of course, you prompt for it 😅).

This appears to be a torch bug, and we'll revert this change once it is resolved.

All Changes

Enhancements

  • Support for FLUX ControlNets.
  • Improved UX for common img2img flows in Canvas.
  • Installer option to select best packages based on GPU model, and internal logic to select the best attention type based on GPU model.
  • Updated workflow list menu UI, restoring New Workflow confirmation dialog.
  • Support for bulk image uploads, button in gallery to upload images.
  • Boards maybe sorted by name or date.
  • Recall FLUX guidance parameter. Thanks @rikublock!
  • Add starter model bundle for each supported architecture to model manager.
  • The Layers and Gallery tabs now remember which one you last selected.
  • Added an indicator to the Layers and Gallery tabs when dragging an image to indicate you that you can hover over the tabs to change them.
  • Updated translations. French is now 100%! Thanks gallegonovato, @Harvester62, @Ery4z!

Fixes

  • Workaround for Apple Silicon SDXL noisy mush issue.
  • Fixed misc UI jank in workflow list menu UI.
  • Fixed longstanding issue where workflows were marked as unsaved immediately after loading.
  • Fixed canvas layer preview not updating if layer is disabled.
  • Fixed an edge case where entity isn't visible until interacting with canvas.

Perf

  • Reworked gallery rendering and context menu for a ~100% perf boost when rendering the gallery.
  • Rendering optimizations for canvas.

Installation and Updating

To install or update to v5.2.0, download the latest installer and follow the installation instructions

To update, select the same installation location. Your user data (images, models, etc) will be retained.

What's Changed

Full Changelog: v5.1.1...v5.2.0

Contributors

weblate, ebr, and 8 other contributors
Assets 5
Loading