High Dynamic Range Sensors, Explained: Camera HDR vs AI HDR

Stand in a cabin at noon and look out the window. You can see the grain of the wooden table in front of you and the texture of the clouds outside at the same time. Now take the shot. One camera keeps the clouds and turns the table into a black slab. A newer camera keeps a bit of both. A phone produces something surprisingly balanced, then applies its own look to it. Same window, same light, three different answers. The difference sits inside the sensor, and it has a name: dynamic range.

Where dynamic range comes from

Every pixel on a sensor is a tiny bucket that collects photons during the exposure. The bucket has a maximum capacity, called full well capacity. Fill it past that point and the extra light spills into nothing: the pixel records pure white, and whatever detail was in that part of the scene is gone. At the other end, the sensor's own electronics add a small amount of noise when the charge is read out. Detail dimmer than this read noise floor drowns in it.

Dynamic range is the distance between those two walls, usually measured in stops. A large pixel with a deep well and clean readout might cover 14 or 15 stops. Cram the same pixel count onto a phone sensor a fraction of the size and each bucket shrinks. Small wells fill quickly in bright light and produce noisier shadows in dim light, which is why a phone struggles with a high contrast scene that a larger camera handles comfortably.

This is physics, not marketing. Sensor makers can improve readout circuitry and microlenses, but they cannot repeal the relationship between pixel area and how much light a pixel can hold.

How modern sensors stretch a single frame

Engineers have found ways to widen that single frame without making the sensor bigger.

One is dual conversion gain. A pixel can read its charge in two modes: a low gain mode that preserves highlights by using the full well, and a high gain mode that amplifies the signal so shadow detail clears the noise floor. Many current sensors switch modes at a specific ISO, which is why some cameras show a second jump in shadow cleanliness partway up the ISO dial. Some designs capture both readouts at once and merge them, extending the range of one exposure.

Another is staggered HDR, common in phone and video sensors. The sensor reads its rows at different times with different exposure lengths: some rows get a short exposure that protects highlights, others a long one that digs into shadows. The processor stitches the row groups into one frame with more range than either exposure could hold alone.

A third approach is in-sensor bracketing, where the sensor or the camera's pipeline fires several exposures in rapid succession and merges them before you ever see a file. It is the classic HDR bracket moved into hardware, fast enough for handheld shots.

All three share one trait. They work at the moment of capture, on light that is hitting the sensor right now.

The second layer: computational HDR on phones

Phones add another stage on top. Features like Smart HDR and HDR+ shoot a burst of frames per shutter press, align them, and merge the best parts of each. This is software bracketing, and it is why a phone with a tiny sensor can produce a sunset photo that embarrasses an older dedicated camera.

Two limits follow. First, the phone decides the look for you. Tone mapping choices are baked in, and aggressive ones produce the flat, slightly grey HDR look that phone photos are known for. Second, and more important for this article: like the sensor techniques above, computational HDR only exists at capture time. It cannot touch a photo that is already taken.

Where hardware hits its limit

That boundary matters more than any spec sheet number. A better sensor helps the next photo you take. It does nothing for the thousands already in your library. The vacation shots from 2014 with blown skies, the phone photos from before multi-frame merging existed, the scanned family prints. No sensor upgrade reaches backward.

RAW files soften the boundary but do not remove it. A RAW file keeps more of what the sensor recorded, so you can lift shadows a stop or two in editing before noise takes over, and recover highlights that a JPEG would have clipped. But a RAW file still holds only what the sensor captured. Data clipped at the well's ceiling is not in there. And most photos in most libraries are JPEGs anyway, with the camera's own tone curve already applied and the extra data discarded.

For the perception side of why single frames lose the ends of the range, our article on why photos look flat goes deeper, and the beginner's guide to HDR photography explains the bracketing workflow that hardware has been busy automating.

The software path: rebuilding range after the fact

The newer option works on the other side of the shutter. A single-image AI model studies an existing photo, estimates what the clipped highlights and crushed shadows most likely contained, and rebuilds the image with a wider tonal range. No bracket, no tripod, no RAW file required.

That is what Free HDR Photo Enhancer does. You upload a JPEG, PNG, or WebP, pick a model and resolution, and get an HDR version back in about 30 to 60 seconds. There is nothing to install; the image is processed on the server. It is free to use (daily check-ins grant credits), and if a generation fails on our side, the credits are refunded automatically.

Two honest caveats. The output is a reconstruction, not a recovery: the model infers plausible detail, so fine textures may differ from what the original scene contained. And camera HDR vs AI HDR is not an either/or choice. Hardware HDR extends what you can capture going forward. AI HDR revisits what you already captured. They solve different halves of the same problem.

What to do with this

If you are shopping for a camera, dynamic range figures are worth reading, but do not buy a new body for range alone. Any camera from the last decade, helped by the in-sensor techniques above, covers most scenes most of the time. Light and timing will improve your photos more than a stop of extra well capacity.

For the photos you already have, the path is shorter. Pick one where the sky went white or the shadows went solid, run it through the enhancer, and compare the two at full size. That single comparison teaches you more about dynamic range than any spec sheet. For a side-by-side look at what the enhanced output actually changes, our HDR vs SDR comparison guide shows the difference in practical terms.