32 bit floating point is a colour depth setting that essentially switches your software from display referred to scene referred.
What is display referred working?
Display referred workflows work with the final output in mind, this could be sRGB (for computers), Rec.709 (for HDTVs), P3 (for cinema or Netflix 4k), and there are many other options. This means that the software is compressing the colour range to fit within the capabilities of the screen it will be shown on.
32 bit floating point essentially creates an infinite colour range and switches you from a display referred workflow to a scene referred workflow (not regarding where the piece will end up, just working in regards to your scene). This is preferable in comparison to 8 bit or 16 bit because it means that you can now use values higher than 1 for white.
In 32 bit floating point, if an object has a value of 1, it is called ‘paperwhite’; it is white but no brighter than a piece of paper. If the value then goes above 1, then the object starts to act as a light, as it is brighter than white.
Another key point to ensure that 32 bit floating point works properly is linearising your workflow, this ensures that the software can simulate infinite stops of dynamic range, rather than limiting it to the 8 stops that sRGB can offer. This means that light starts to act as it does in real life, and brings what the software creates to be more realistic like what we can see with our human eye.
Playing with light in After Effects
I decided to switch my After Effects to 32 bit floating point from 8 bit, and then experimented with creating colours using values higher than 1. I found that if you have two values above 1, at different levels, you can get some nice gradient effects. When I exported it however I forgot to change the colour depth settings in the export and ended up losing the bright white that I had had in the viewport. I still think it looks really nice though and will have to keep experimenting with settings to see what I can add to my work.
Another minor tool I learned about through this experiment was the ability to change the shutter angle. This creates more ‘inbetweens’ and a smoother motion blur, which I really liked.
Possibly one of the big issues for me with 32 bit floating point is that my low powered computer doesn’t handle it well. Rendering previews into the RAM took a lot longer than usual considering the simplicity of my scene, however, I think that I could definitely work around it, or just put up with the slower workflow for the benefits that 32 bit floating point brings.
Jennie Hawkes 