Update text_detection.py

Update text_detection.py

Preserve only text mask regions under 50 pixels tall

Show character-shaped text mask overlay in ignored preview

Preload GPU models at startup and reuse them for inference

Add ignored-region bounding box preview beside cleaned output

Preserve segmented text characters with two-pixel padding

Isolate PaddleOCR from Gradio and limit CPU threads

Avoid Paddle CUDA dependency conflict with PyTorch

Mask preserved regions during diffusion and ignore text by default

Update app.py

Update requirements.txt

Update README.md

Simplify ignore checkbox labels

Use HyperHail bucket for Archivist Soft cleanup model

Keep processing output resolution when restoring ignored regions

Add final 1x cleanup and full-resolution Gradio image output

Add anime hand and foot preservation options

Preserve detected eyes during watermark removal

Prefetch every model before app launch

Defer CtrlRegen asset fetch until first use

Fetch only selected model assets

Avoid zero-step inference and prefetch Illustrious assets

Merge remote GPU duration change

Limit GPU scope to model inference

Download default and CtrlRegen assets at startup and export PNG

Update app.py

10

Make CtrlRegen the default selection

Let upload previews fill their rows

Resolve Illustrious checkpoint by repository filename

Enlarge upload previews

Use Illustrious XL for default ControlNet regeneration

Add SDXL anime regeneration option

Show fixed-size previews for uploaded images

Add default Super Minimal preset

Download cleaned image in one click

Preload CtrlRegen and add default-on toggle

gitignroe

Allocate 10s ZeroGPU for preloaded watermark inference

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py

Update app.py