AutoGain

AutoGain analyzes the incoming image automatically and finds its current extreme values. The darkest value maps to Low, the brightest value maps to High, and intermediate values are redistributed linearly between them. With the default Low 0.0 / High 1.0, the useful range is stretched from black to white.

Low and High must be read as target output Range endpoints. They are not input clipping thresholds and they do not manually choose which source values count as black and white. AutoGain chooses the input extrema itself.

Blackmagic documents a simple reference test: feed a horizontal gradient from 0.2 to 0.8, then set Low to 0.0 and High to 0.5. The darkest end lands at 0.0, the brightest end lands at 0.5, and the remaining values scale between those endpoints.

Automatic extrema detection is both the convenience and the temporal hazard. If a bright object leaves an otherwise dark shot, the new maximum becomes lower and the remaining scene is stretched more aggressively. In playback this appears as global brightness/contrast pumping even though no AutoGain parameter is animated.

The same issue exists at the dark end. A very dark object, shadow, clipped region, or technical outlier entering or leaving can change the detected minimum and shift the rest of the range. A single hero frame is therefore not evidence of temporal stability.

Do Z applies the AutoGain operation to Z/Depth channels. Blackmagic describes this as useful for matching the ranges of Z channels or making floating-point depth easier to inspect. It does not generate depth; the upstream image must already contain valid Z data.

Do Z needs special care in production. If a near or far object enters or leaves the scene, the automatically discovered depth range changes and existing objects can receive different normalized Z values on the next frame. That can be destructive for downstream depth-dependent work even when the Viewer looks easier to read.

For that reason, AutoGain with Do Z is best treated as a diagnostic or controlled range-matching branch rather than mandatory preparation for every depth pass. The Resolve manual also notes that the Viewer's normalized-image option is generally more suitable when the task is simply to inspect the full range of floating-point imagery because it does not have to become part of the processed image path.

Effect Mask is a separate blue input. Resolve 21 explicitly states that the mask is applied after the tool has processed the image and limits the final AutoGain adjustment spatially. That means it should not be documented as a guaranteed selector for the statistics used to discover the extrema.

The post-process mask semantics matter in practice. A mask can restrict where the AutoGain result contributes, but the automatic normalization remains sensitive to the input frame. Do not build a workflow around the assumption that a Rectangle makes AutoGain 'measure only the face' unless that exact behavior has been verified live in the target build.

AutoGain is valuable for quick technical work: expanding a narrow grayscale pass, checking whether floating-point imagery contains meaningful range, making low-contrast diagnostic material easier to read, or comparing controlled depth passes. Creative exposure, contrast, and shot matching are usually better handled by tools with explicit stable parameters.

A very bright specular, hot pixel, noise spike, or isolated extremely dark value can become an extremum and force most of the image into a small portion of the target Range. If the result unexpectedly looks flat, investigate outliers before simply raising High.

HDR and float workflows can contain values far outside 0–1. AutoGain can remap that range into something easier to inspect, but it still changes data. Keep Viewer normalization conceptually separate from an actual pixel/depth remap in the composition.

Low/High do not have to be 0–1. A High below 1.0 preserves upper headroom in the output representation; a Low above 0.0 raises the lower endpoint. The operation remains the same: current input min/max map to the chosen output endpoints.

Common Settings such as Blend and channel-processing controls belong to Fusion's shared Settings infrastructure and are not additional dedicated AutoGain controls. The captured Controls panel for this article contains only Do Z and the two Range values, Low and High.

The main production test is frame-to-frame continuity. Scrub and play through the largest lighting changes, appearing/disappearing highlights, and—when Do Z is enabled—near/far geometry changes. If global levels or depth values drift, replace automatic normalization with a more stable explicit pipeline.