●Change8

v6.0.0

Breaking Changes
📦 invokeaiView on GitHub →
⚠ 2 breaking✨ 15 features🐛 11 fixes🔧 3 symbols

Summary

Invoke 6.0 introduces a major UI overhaul with customizable layouts, new Launchpads, and enhanced Canvas features like PSD export and FLUX Kontext integration. This release also brings significant performance improvements and addresses numerous bugs.

⚠️ Breaking Changes

  • The 5.0 layout has been retired and replaced with a new task-specific, flexible, and persistent interface foundation. Users must adapt to the new UI structure.
  • The Infinite Scroll feature has returned, now showing all images at once in a single grid, replacing previous pagination methods. This change may disrupt workflows relying on simple chunk-based navigation.

Migration Steps

  1. Review the new UI layout, as the 5.0 layout has been retired.
  2. If relying on pagination for image viewing, adapt workflows to the new Infinite Scroll grid view.
  3. Users on older GPUs (2XXX series) may encounter issues due to recent pytorch updates; manual downgrading of torch may be necessary. Seek help on Discord if needed.
  4. Follow the Quick Start guide using the Invoke Launcher for the recommended installation/update process.

✨ New Features

  • Introduction of 'Launchpads' in Generate, Upscale, Workflow, and Model tabs to guide new users.
  • Customizable layout allowing users to drag, drop, and resize panels on each tab; custom layouts are saved automatically.
  • Global reference image management consolidated into an intuitive zone in the left-hand settings panel.
  • Tabs now load on demand for improved performance and responsiveness.
  • Rule of Thirds Guide added to Canvas settings for composition assistance.
  • New button to instantly create a reference image from the contents of a bounding box on the Canvas.
  • New 'Edit' button in the image viewer to create a new canvas with the selected image.
  • New 'Control Layer Resized' drop zone for automatically optimizing control layers when added to an existing canvas.
  • Overlay Control feature to toggle the visibility of all non-raster layers (Control Layers, Inpaint Masks) via a button or Shift+H hotkey.
  • Improved staging toolbar and navigation on Canvas, with preview image controls directly on the toolbar.
  • New option to automatically save every image generated on the Canvas to the gallery.
  • Canvas Export to PSD (.psd) functionality, preserving all raster layers.
  • Full UI integration for FLUX Kontext Dev, supporting reference image guidance for dev and quantized (gguf) variants (fp8 not supported).
  • LoRA Picker Overhaul: Uses the new model picker, automatically filters LoRAs compatible with the selected base model, and supports preview images/related models configuration.
  • Support for OMI LoRA and additional AI Toolkit LoRA for FLUX.

🐛 Bug Fixes

  • Fixed names of unpublishable nodes.
  • Ensured each invocation overrides _original_model_fields with its own field data.
  • Fixed 1:1 ratio for flux kontext.
  • Fixed regional guidance reference image not selecting correctly in the UI.
  • Fixed LoRA picker to default to the current base model architecture.
  • Fixed UX when waiting for an image.
  • Prevented NaN from entering Konva internals.
  • Fixed preparation of collect nodes which could create invalid edges.
  • Fixed related models display in the new model picker.
  • Fixed Upscale creativity bug.
  • Fixed navigation on Edit/View workflow buttons to point to the correct panel.

🔧 Affected Symbols

flux_text_encoder.pyflux_denoise.pyFieldIdentifier