ADVANCED COMPOSITING & RETOUCHING - Chapter 2, Exercise 1 Why You Must Duplicate a Channel Before Adjusting It ==================================================================================== QUESTION: Explain, step by step, why you should duplicate a color channel before adjusting its contrast for a mask, and what specifically goes wrong if you skip that step. SOLUTION / EXPLANATION: An RGB image's Red, Green, and Blue channels aren't just a preview or a diagnostic view - they ARE the actual color data of the image. When you view the "Red channel" in the Channels panel, you're looking at (and, if you edit it, directly editing) the real red component of every pixel in the photo. Step by step, what goes wrong if you skip duplicating: 1. You select the Red channel (for example) because it shows the best contrast between subject and background. 2. You apply a Levels or Curves adjustment to push the subject toward white and the background toward black - but you've applied this directly to the image's real red channel data, not a copy. 3. The canvas still displays as a single-channel grayscale view while you're inspecting that channel, so nothing looks obviously "broken" yet. 4. Once you finish building your mask and switch back to the full-color composite view (viewing all three channels together again), the image's actual color is now wrong - the red channel has been contrast-boosted in a way that has nothing to do with how the photo should actually look, and there's no way to separate "the mask-building adjustment" back out from "the real color data" once it's applied. 5. The only fix at that point is undoing all the way back past the mistake, or reverting to a saved copy - the edit can't be selectively removed after the fact. Duplicating the channel first (most programs offer a simple "Duplicate Channel" command in the Channels panel menu) creates a completely separate grayscale layer that behaves like any other image - free to adjust with Levels, Curves, painting, or Threshold as aggressively as needed, with zero risk to the original RGB data, since the duplicate isn't linked back to the real channels at all. Once the mask work is finished, the duplicate is loaded as a selection and then can simply be discarded. -------------------------------------------------------------------------- WHY THIS WORKS AS AN ANSWER: It explains why channels are live image data rather than a disposable preview, walks through the specific failure sequence (the mistake being invisible until switching back to full color), and names why the damage is effectively unrecoverable without undo/reverting - then explains what duplicating actually protects against.