Browse or paste an image to neutralize tan/sepia casts while lifting contrast with protected highlights.
White Reference Depth chooses the bright color value the correction treats as the image's white reference. The number is a percentile of the visible pixels, measured separately in red, green, and blue after the app picks a shared 2% black point. Around the default 99.45%, each channel samples near the bright tail instead of the absolute maximum, then builds a separate channel gain that maps that anchor to a calculated light target. Lower values pull the reference deeper into the image's color mass instead of trusting only rare highlights, which usually removes tan or sepia casts more strongly and opens midtones. Higher values are more conservative, preserving warm lighting and dark atmosphere, but may leave more of the original cast. The highest highlights are protected later by a soft-knee compressor, so this control is mainly about which part of the distribution is trusted for white balance.
White Target Lift shifts the brightness target used for that white reference. The zero point uses the automatically calculated target without extra lift. Positive values push the reference higher in the output scale, sometimes beyond display white; that increases contrast, makes whites cleaner, and lets the soft highlight knee roll off excess brightness. Negative values pull the target down for a denser, safer image with less risk of washed highlights. Because this changes the per-channel normalization anchor rather than adding a flat exposure offset, it can also change how neutral the midtones and highlights feel.
Simple mode restores the earlier single-slider compressed normalization. Highlight Clipping chooses how much of the bright tail is treated as clipped when finding each channel's white point. The simple algorithm uses a shared 2% black point, stretches each channel to a 235 highlight knee, and compresses remaining highlights up to display white.