8-bit vs 32-bit Floating Point

8-bit vs 32-bit Floating Point

Bit depth determines how much information each pixel channel can store. In 8-bit, each channel has 256 possible values, from 0 to 255. That seemed like plenty when computers had limited memory, but it creates real problems for VFX work. If you try to brighten an 8-bit image, the gaps between those 256 values become visible as ugly bands instead of smooth gradients. This is called banding, and it is a nightmare to fix.

16-bit gives you 65536 values per channel, which is much smoother. Gradients look cleaner, and you have more room to adjust exposure and color before quality degrades. Most intermediate VFX work uses 16-bit integer, and it is a good balance between quality and file size. But 16-bit still has limits. If you push it too far, you will still get banding, and it cannot represent values brighter than pure white.

32-bit floating point is a different beast. Instead of storing integer values, it uses floating point numbers with a mantissa and exponent, just like scientific notation. This gives you an enormous range of values. You can represent 0.001 just as easily as 10000. This is critically important for HDR rendering because light in the real world does not stop at white. A light source can be ten times brighter than a white piece of paper, and you need floating point to store that correctly.

When you do color corrections and compositing operations, 32-bit floating point preserves all the precision of your calculations. If you multiply two values together, you get a precise result instead of rounding errors. This matters a lot for effects like glow, motion blur, and depth of field that need accurate color math. Working in 8-bit for these operations will give you wrong results that are immediately obvious to the trained eye.

In a practical pipeline, you use different bit depths for different stages. Shoot and render in 32-bit float or 16-bit for maximum quality. Do your compositing in 32-bit float to preserve precision. Only convert to 8-bit at the very end for delivery formats like MP4 or web video. This way you keep all the quality through the creative process and only lose what is necessary for the final distribution.