2.5.1

• Bug fixes

2.5.0

• New UI
• LoRA support during generation. ( Now you can use multiple LoRAs without merging)
• Seed compatibility mode ( to reproduce results from A1111 etc )
• Support for bfloat16 in LoRA
• New feature to create illusions

2.4.4

• Improvement in SDXL quality
• Prompt weighting like A1111
• Text prompt styles
• New default model

2.4.3

• Bug fixes
• Support for training custom SD models locally

2.3.0

• 30% speed improvement
• Support for SDXL

2.2.1 ( Beta )

New features

• Completely new UI
• Ability to queue generations
• Video tools - interpolation, Deforum
• A new infinite AI canvas
• Ability to draw on canvas
• More controlnet models
• Better in-painting and outpainting
• LoRA tools
• More advanced options for generation

1.7.4

New features

• ControlNet
• SD2.x support
• Fixes black screen issue with Intel
• Ability to generate similar images
• Multiple schedulers
• Safetensors support

1.5.2

New Features

• Safer way to import custom models
• Some bug fixes
• Custom models option in front
• Resizable outpainting box
• Negative prompt for outpainting

1.5.1

New Features:

• Ability to add and select custom DreamBooth models
• New options menu
• Sound on generation
• Toast notifications
• Pagination in history

FP16 version runs much faster, but has slightly worse generation quality. FP32 version produces the best quality.

Warning : loading external custom models can potentially harm your computer. So please only load models from trusted sources.

1.4.3

Features:

• Better outpainting UI
• Bug fixes
• TF model updated

Notes

• DiffusionBee-1.4.3-arm64_MPS_SD1.5_FP16 runs much faster, but has slightly worse generation quality.
• DiffusionBee-1.4.3-arm64_TF_SD1.4_FP32 runs slower, but produces better quality images