diffusionbee-stable-diffusion-ui

• Support for Flux.1 image generation models ( only for arm64 and MacOS 13+ )
  • To use flux, scroll down to the bottom in the app home screen
• Support for external textual inversion embeddings
• Ability to block NSFW images
• More organized models page
• Several bugs fixed

• Ability to import SD models from Hugging Face
• Issues in downloading the main model fixed

• Bug fixes

• 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

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

• Bug fixes
• Support for training custom SD models locally

• 30% speed improvement
• Support for SDXL

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

New features

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

New Features

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

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.

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

Features:

• Outpainting
• Stable Diffusion model v1.5 added
• Separate UI for inpainting

Note: The intel version is experimental and does not work with AMD GPUs

Features:

• UI fixes
• Inpainting
• Negative Prompts

Features

• Sending images to img2img
• Upscaling using RealESRGAN
• Option to share on arthub.ai
• fixed border issue for img2img
• selecting prompts / params in history

Features:

• Flexible size for image to image
• Bug fixes

New Features

• Image to Image
• Dark theme ( Thanks to Iktwo )
• UI Improvement

New Features:

• UI Fixes
• Persistent history
• Bug fixes

New Features :

• Batch Size
• Fast runtime on MacOS Ventura
• Bug fixes
• Image viewing options.

New features:

• Option to set seed
• Temporary fix for blank images in MacOS Ventura