social media
Why Social Video Looks Worse Than You Shot It: Compression, Banding and Bitrate
Why social video looks worse than you shot it: how re-encoding causes banding and blocking, and what to do in the studio, from lighting to a test upload.

You graded a smooth coloured wash on a monitor, exported a clean file, uploaded it, and the version you see on the phone has steps in the backdrop and mush in the shadows. Nothing went wrong in the edit. The video was re-encoded on its way, and a few kinds of picture suffer more than others.
Most of those kinds are decided on set. This guide explains the physics in plain terms and then what to do about it in the studio.
What the platform does to your file
Platforms re-encode what you upload, so assume yours does. The copy a viewer sees is a second-generation file, made smaller so it plays on a phone on whatever connection it is on. Lossy compression works by deciding which detail the eye is least likely to miss and throwing it away. It is good at that most of the time, and the picture you made is the one that tests it.
How hard the file is squeezed is the platform's choice, and it changes. Check the platform's current spec for what it accepts, and do not build your plan on a figure you read last year.
Why gradients band
Most delivered video is 8-bit, which gives 256 levels per colour channel. A smooth gradient from one shade to a nearby one only spans a handful of those levels, and the encoder has to describe each patch with few bits. Neighbouring areas get rounded to the same value, and the smooth fade becomes visible steps. That is banding.
Dark scenes make it worse. In the shadows each step between levels is a larger fraction of the brightness, the eye is very sensitive to steps there, and low-contrast shadow detail is among the first things a lossy encoder gives up. A near-black gradient is the worst case.

Where else it breaks
- Fine texture: knitwear, hair, glitter, a patterned shirt and film-style grain cost the most bits, and the encoder smears them or turns them into blocks.
- Heavy motion: when every pixel changes between frames the encoder cannot predict one from the last, and blocks appear in the moving areas.
- Noise: grain from a high sensitivity setting is texture the encoder spends bits on and then fails to keep, so it eats into the quality of the detail you wanted.
- Washes and vignettes: a saturated cove wash that darkens towards the corners is a smooth gradient in colour, which bands.
What this means on a studio set
The cove is the trap. A coloured wash on Studio B's cyclorama, or any white cove that falls off to grey at the edges, is exactly a smooth gradient. So is the pool of light on a black backdrop and the soft falloff you work to make a subject look dimensional. They look right in the room and on the grading monitor, because the monitor has the whole file. The viewer's copy does not.
That is not a reason to avoid colour. It is a reason to design the backdrop for the second-generation file, not for the monitor.
Countermeasures you can use on the day
- Keep washes simple. One flat, even colour holds up better than a blend of two with a gradient between them. The cyclorama guide shows how to get a wash even.
- Add slight texture. A backdrop with a little surface, a pattern of light across it, or some objects in the background gives the encoder something to hold on to and hides the steps; a fine, even grain added in the grade helps for the same reason, which is not the same as high-ISO noise from underlighting.
- Avoid near-black gradients. Light the backdrop enough that it reads as a colour, or go to a flat dark and keep the fall-off out of the frame.
- Light so the subject separates from the backdrop. A rim or a side light and a clear difference in tone mean a soft backdrop can be simple.
- Shoot clean. Low noise needs enough light, and lights are charged separately from the room rate in every room but the DJ room — please enquire about our lights and we will quote for what your shoot needs. The studio lighting fundamentals guide has the basics.
- Slow the motion. A locked-off or slow-moving shot compresses better than a whip pan.
Camera settings help too: keep the sensitivity low, avoid sharpening in camera, and follow the camera settings guide.
Export high and test before the day ends
Export at the highest quality the platform accepts, at the resolution and frame rate it asks for. Do not compress the file yourself first, because the platform will compress it again and you pay for two generations of loss. Check the platform's current spec for the file types it takes.
Then test it, as the next section sets out. Shoot a short clip on set, with the real backdrop, the real light and the real wardrobe, and upload it as a private or draft post, where the platform allows, before the day is over. Look at it on a phone, in the app, not on the editing monitor. If the wash is banding, you can fix it while the rig is still standing, in a booking where set-up and pack-down sit inside the time. The building has Wi-Fi, but its speed is not published, so ask before you rely on it for a test upload.
What to look for in the test clip
- The backdrop: are there visible steps in the wash, or in the falloff behind the subject?
- The shadows: does the dark side of a face, or a dark garment, turn to blocks?
- Texture: do the patterned wardrobe, hair and props hold their detail?
- Movement: does a hand moving through the frame smear the background?
- Captions and text: are they still legible at phone size and clear of the top and bottom edges?
Design the backdrop for the second-generation file. Shoot a test clip and look at it on a phone before the rig comes down.
The vertical-frame craft that sits alongside this is in shooting paid social video ads in a studio, and the content day method is how to fit a test into the running order.