Actual Pixels, Don’t Resize - If checked the program saves cropped images without any scaling.
Set Width to - Here you may select the width of exerted images in pixels. The program will scale exported images to that size.
Don’t enlarge smaller images - If cropped area is less than required export size it will not be up-scaled
If Oversized Crop - Shrink To Fit Image Bounds - Program calculates crop borders automatically with regard of face size and position. It can happens that face on the photo is too big or located close to the image edge. In this case the automatic crop box will be bigger than image itself.
If you select Shrink To Fit - the cropping box will be resized to optimal position within image bounds keeping face centre in the position required by preset.
If this option is not checked the program expands cropping border outside the image. Outside areas are filled with color the same as top-right point of the cropped rectangle. If you shot your models over plain color background (like green screen for future background removal) you’ll get quite good results suitable for further work. Otherwise you may want to disable this option to avoid visible color margins.
The settings above can be also set for any preset individually. They are named Export Width and Oversize. Preset settings override default preferences setting.
JPEG Quality (1-100) - This is jpeg compression. 100 means full quality. 1 - very compressed with visible artefacts. In real life numbers about 80 - 90 give you quite good pictures.
Auto save batch info to faf.autosave file - The program saves information about opened files, detected face positions, and cropped areas, including manually adjusted crops into faf.autosave file located in currently opened folder. When you’ll work with this folder again, you may open faf.autosave (instead of opening individual images) and instantly get all faces detected and cropping boxed positioned as it was in last working session.
Face Detector - Select default face detector here. DUN detector gives you fast and accurate detections for individual or small group photos. HAAR not as fast and accurate as DUN but can be useful to detect many faces on group