- Added Style Iterator.
- Removed meta tensor usage.
- Fixed loading TAESD decoder for SD from custom path.
- Added Load Last Prompt button (contribution from sngazm).
- Fixed hangup in Upscale (Fast 2x).
- Fixed problems with turning off FreeU in some scenarios.
- Fixed loading prompts from JPG files processed by external apps.
- Fixed fast upscale always saving as PNG.
- Added limited support for non-SDXL models (no refiner, Control-LoRAs, Revision, inpainting, outpainting).
- Added support for FreeU.
- Updated Comfy.
- Fixed reading paths from paths.json (broken in 2.0.73 MRE).
- Fixed error related to playing audio notification.
- Fixed error related to loading prompt with Enhance Image mode active.
- Disabling Image-2-Image when Enhance Image is active.
- Added information about total execution time.
- Enforced 'dpmpp_fooocus_2m_sde_inpaint_seamless' sampler for inpainting workflow.
- Updated Comfy and patched Comfy functions.
- Renamed Input Image from vanilla to Enhance Image (to avoid confusion with Input tab)
- Added support for wildcards (ported from RuinedFooocus, adjusted to Fooocus V2).
- Added support for ddpm sampler.
- Restored saving information about real prompt in metadata and log file (adjusted to Fooocus V2).
- Fixed links to log files not working with customized outputs path.
- Disabled Fooocus Virtual Memory from vanilla (not compatible with current Comfy).
- Updated Comfy.
- Added support for authentication in --share mode (via auth.json).
- Added Image-2-Image Scale slider.
- Displaying Revision and Control-LoRAs controls only when needed.
- Added support for loading models from subfolders (ported from RuinedFooocus).
- Updated Comfy.
- Added support for higher resolutions in Image-2-Image mode (can be used for upscaling).
- Updated Comfy (CLIP Vision optimizations).
- Changed Prompt Expansion (aka Fooocus V2) to be enabled by default.
- Moved links to Settings tab.
- Updated Comfy (fixes CLIP issue).
- Restored allowed random seed range (entropy reduction applied only to transformers / numpy related calls).
- Added support for adjusting text prompt strengths (useful in Revision mode).
- Reduced allowed random seed range to match limits in Prompt Expansion and transformers (trainer_utils.py).
- Updated Comfy.
- Renamed Raw Mode (enabled by default in vanilla) to Prompt Expansion (disabled by default in MRE).
- Added support for reading styles from JSON files.
- Added support for playing audio when generation is finished (ported from SD web UI).
- Fixed joining negative prompts.
- Increased Control-LoRAs strength range.
- Allowed passing parameters to Comfy.
- Added links in Misc tab
- Updated Comfy.
- Added support for custom resolutions and custom resolutions list.
- Updated Comfy.
- Added MRE changelog.
- Added support for Control-LoRA: Depth.
- Added Canny and Depth model selection.
- Added ability to stop image generation.
- Added support for generate forever mode (ported from SD web UI).
- Added information about prompt execution time.
- Changed default Control-LoRA rank to 128 (allows using both on 8 GB VRAM GPUs).
- Fixed problems with random seed precision.
- Increased maximum number of generated images to 128.
- Moved Revision in front of Control-LoRAs.
- Updated Comfy.
- Updated Gradio.
- Updated Colab notebook.
- Added support for Revision (thx to comfyanonymous for helping figuring out issues related to lllyasviel changes).
- Added support for Control-LoRA: Canny.
- Added support for Embeddings.
- Added support for JPEGs (with metadata saving & loading).
- Added support for generating multiple images using same seed (useful in Image-2-Image mode).
- Added support for keeping input files names.
- Added information about aspect ratio to resolution list.
- Added dpmpp_2m sampler.
- Automatically import images into Image-2-Image and Revision modes (if available).
- Automatically set batch size to number of loaded input images.
- Loading only base model on start.
- Updated Comfy.
- Updated Colab notebook.
- Added support for Image-2-Image mode.
- Added support for changing paths (via paths.json).
- Added Custom performance mode.
- Fixed prompt loading.
- Implemented cycling over input images in Image-2-Image mode.
- Made resolution and style selection more compact (thx to runew0lf for hints).
- Simplified initial view.
- Updated Comfy.
- Improvemed VRAM management.
- Corrected Colab notebook.
- Updated Comfy (+DISABLE_SMART_MEMORY).
- Fixed problem with disabled refiner (by youcheng).
- Layout corrections.
- Updated documentation.
- Renamed fork to "Fooocus-MRE".
- Added support for viewing metadata in the UI.
- Added support for loading basic UI settings from file (settings.json).
- Allowed saving metadata in both JSON and PNG file.
- Fixed problem with handling text files encoding (for non-ASCII prompts).
- Added support for loading prompt parameters from JSON and PNG files.
- Added support for changing scheduler (karras, exponential, simple, ddim_uniform).
- Enhanced list of available samplers (added dpmpp_sde_gpu, dpmpp_sde, dpmpp_2s_ancestral, euler, euler_ancestral, heun, dpm_2, dpm_2_ancestral).
- Updated Comfy.
- Added support for dpmpp_3m_sde_gpu, dpmpp_3m_sde, and dpmpp_2m_sde samplers.
- Added support for saving prompt information in PNG metadata or JSON file.
- Added support for changing steps, switch step, and CFG.
- Added support for changing CLIP Skip values.
- Changed resolution list to official SDXL resolutions.
- Enhanced RNG seed range.