Anyone who has watched an afternoon match in a partly covered stadium knows the picture: one half of the pitch bleached by sunlight, the other half sunk in shade, and players vanishing as they run from one into the other. We cover the problem itself in Shadows in live sports. This article looks at one family of fixes, real-time tone mapping, and at why the way it is applied matters as much as the idea itself.
What tone mapping does, and why a fixed LUT can't follow a moving shadow
A sunny stadium can hold far more contrast than a broadcast signal or a living-room screen can reproduce. Tone mapping is the general name for the techniques that compress that range: deciding how bright highlights should be rolled off, how much detail to pull up from the darks, and how to keep mid-tones (skin, shirts, grass) looking natural in between.
The most common tool in a live chain is the look-up table, or LUT. A LUT is a fixed mapping: a given input value always becomes the same output value. That is exactly what you want for converting between colour spaces or applying a consistent house look. It is also its limitation. A LUT has no idea where a pixel sits in the frame. It cannot tell a dark shirt in sunlight from a white shirt in shadow if both arrive at similar values, and it cannot treat the shaded half of the pitch differently from the sunlit half.
Shadows in live sport also move. The shadow line creeps across the pitch through the afternoon, cameras pan and zoom, and the balance of sun and shade in any given shot changes from second to second. A global curve tuned for one shot is wrong for the next. Vision engineers compensate by riding iris and gain, but that still applies one setting to the whole frame: you protect the highlights or you open up the shadows, rarely both. We explain that trade-off in more depth in Exposing for sun and shade.
Frame-by-frame, region-aware enhancement
The alternative is to stop treating the frame as one thing. Region-aware enhancement looks at each frame, works out which areas are in direct sun and which are in shade, and adjusts them separately, so that a player crossing the shadow line stays readable on both sides of it. Because the analysis is repeated on every frame, it follows the shadow as it moves rather than relying on a setting chosen at kick-off.
This is where LivePictureAI fits. It is not a tone mapper in the classic sense and it does not replace your LUTs or your grade. It processes the feed frame by frame to rebalance sunlit and shaded regions, lifting shadow detail and taming highlight glare, so that the picture moving down the chain already has a usable balance between sun and shade. Not an effect. Not a filter. The aim is a picture that looks the way the stadium looked to someone sitting in it, not a stylised one.
The important distinction from a shadow "eraser" is that the shadow stays. Viewers expect to see that part of the pitch is in shade; what they should not lose is the ball, the players' numbers or the run of play. For more on that difference, see AI shadow removal for live video.
SDR and HDR delivery
Across the industry, many productions now carry more than one output: an HDR programme for some distributors and an SDR version for everyone else, often derived from the same cameras. In that world, tone mapping turns up in several places, typically when converting between HDR and SDR, when matching cameras with different characteristics, and when making a single production look right on very different displays.
Wider-range delivery does not make the sun-and-shade problem go away. A brighter, wider signal can carry more of the scene, but the viewer still needs the shaded half of the pitch to be legible, and most viewers still watch in SDR. Whatever the delivery format, the underlying job is the same: decide what matters in each part of the frame and make sure it survives the trip to the screen. Any frame-by-frame processing in the chain has to sit comfortably alongside the colour pipeline a production already runs, which is why LivePictureAI sits on the camera feeds or on the programme output and leaves the grade and LUTs to the production.
Latency and consistency requirements for live
Offline, tone mapping can take as long as it needs. A real-time tone mapping broadcast setup has hard constraints that offline work does not.
- Latency. Every stage in the chain adds delay, and live production has a strict budget for it. LivePictureAI adds at most 100 ms of latency, designed to keep pace with live broadcast timing.
- Temporal consistency. Adaptive processing that reacts too sharply produces pumping and flicker, especially when a camera pans across the shadow line or a cloud passes over. Changes have to be smooth from frame to frame, so the correction is invisible as a process and only visible as a better picture.
- Fitting the chain. The processing has to accept the signals a production already uses. LivePictureAI works with SMPTE 2110, NDI, SDI, RTSP and file-based streams, and runs on GPU infrastructure at the edge or in a production hub, on-premises or in the cloud.
None of this replaces the vision engineer. Camera matching, exposure decisions and overall picture quality remain theirs; the aim is to take the constant sun-and-shadow balancing off their hands.
AI contrast enhancement for live video
"AI contrast enhancement" and "tone mapping" are often used interchangeably in searches, and they overlap. The useful question is not the label but what the process actually adapts to. A fixed curve adapts to nothing. A global auto-exposure adapts to the whole frame at once. Region-aware, frame-by-frame processing adapts to where the light actually falls in each shot, which is what a half-sun, half-shade pitch demands.
We have tested LivePictureAI on football footage, where the shadow line is most visible and most disruptive, and we are validating it with selected partners after IBC 2026. If you run outdoor sport and recognise the problem, thebroadcaster overview shows where it would sit in your chain, and we would be glad to look at your footage with you.