Certain things are difficult to compress. Here’s a list of things to reduce or eliminate when filming.
The biggest factor in compression is finding a block of pixels that has moved, but this does not account for rotation. The encoder will come up empty-handed when searching for the new location of each block of pixels, causing it to recompress them.
Encoders don’t search the entire image for moved pixels. A common search radius is 16 pixels. Something moving more than this in a single video frame will be recompressed.
Tip: if there is a fast-moving background, use a lens that causes the background to appear blurry. It will be recompressed, but will need less space to do so.
Avoid pinstripe shirts, small polka-dots, and shimmery or glittery clothing. In general, detailed, high-contrast patterns will be difficult to compress.
Cutting to a new scene is better for compression. A fade causes a prolonged spike in bit rate, and will result in lower quality when bit rate must be limited.
If a fade is a must, scenes with less detail are better – solid colors, no patterns, even a slight blur effect will help the quality.
Text without a solid background is difficult to compress. Try adding low-detail backgrounds which are completely opaque.
Graphic and text overlays are unavoidable sometimes, but if you know what resolution your video will be encoded as, start and end your (ideally rectangular) overlay on pixels that are a multiple of 16 for H.264/AVC and multiples of 64 for H.265/HEVC. For example, if you start your overlay rectangle 32 pixels from the left and 32 pixels from the top and it’s width and height are also multiples of 16, it will compress better. Note, the pixel count starts at 0, with the top left pixel of the video being (0, 0).
Zooming also causes a prolonged spike in bit rate. A better option for compression purposes is switching to a different camera.
Explosions, rippling water, shiny object reflecting light, “tv static” effects.