●Change8

v5.9.0rc2

📦 invokeaiView on GitHub →
✨ 4 features🐛 7 fixes🔧 8 symbols

Summary

This release introduces major features including FLUX Fill support for high-quality inpainting/outpainting and support for the LLaVA OneVision VLLM multimodal model. It also includes several bug fixes and developer enhancements like custom string field dropdowns in the Workflow Builder.

Migration Steps

  1. To use FLUX Fill, download it from Starter Models and select it from the main model drop-down on Canvas. Note that FLUX Fill is not compatible with standard Text to Image or Image to Image operations on Canvas (it requires an inpaint mask or empty regions in a bbox).
  2. The new Invoke Launcher is the recommended way to install, update, and run Invoke. Follow the Quick Start guide for launcher installation.
  3. If performing a headless install, follow the manual install guide.

✨ New Features

  • Added FLUX Fill support in Workflows and Canvas, which provides high quality inpainting and outpainting.
  • Added support for the LLaVA OneVision VLLM family of models for multimodal text/image generation.
  • Support for custom string field drop-downs in Workflow Builder, allowing users to define custom options for string fields.
  • Support hot reload for custom nodes (developer-focused change).

🐛 Bug Fixes

  • Fixed issue with installing certain quantizations of FLUX Dev and Schnell (introduced in previous RC).
  • Fixed issue with inpainting graphs that caused bad blending (introduced in previous RC).
  • De-wonkified LoRA node names (names were incorrectly modified in v5.8.0).
  • Fixed issue with Image Viewer where the image could overflow the viewer.
  • Fixed issue where shift-clicking to draw on Canvas ignored the `Clip to Bbox` setting.
  • Improved UX for missing or unexpected fields in Workflows.
  • Fixed SigLip model translation key issue.

🔧 Affected Symbols

FLUX FillLLaVA OneVision VLLMWorkflow BuilderCanvasImage ViewerLoRA nodepicklescanLLaVA OneVision VLLM node