H.264

H.264 is a compression format that dramatically reduces the size of video files. Raw footage straight out of a camera or generated by AI is enormous, so it has to be trimmed down before it can be streamed or sent over the internet. H.264 does this by stripping out the details that are least noticeable to the eye — similar to saving a large photo at lower quality so it's easier to send over a messaging app.

The catch is that color information gets cut down along with everything else. H.264 video can only represent 256 levels of brightness per channel (8 bits), so scenes with gradual color shifts — like a sky — can show visible banding the moment you push color grading even a little. That's much less headroom than formats capable of storing far richer color (10 bits or more).

That's why productions that need heavy color work later, like broadcast or film, don't work directly with H.264 — they convert to formats that hold onto more color information first. Since most generative video AI output also comes out as H.264, a separate class of tools has emerged just to convert that footage into something usable for post-production.

The article explains that "until now, the output of generative video models has been heavily compressed 8-bit H.264 files, so even a small push in color grading would cause banding in the sky and crush the shadows." A common misconception is that H.264 itself is simply a low-quality format — but more precisely, it's a compression method that strips out color data to save space, and that reduces both original image quality and how much room you have to edit later. It's also often confused with concepts like HDR or color space, but H.264 is a compression format, while HDR and color space are separate concepts entirely.