16bit vs 32bit float?

[Brendan Coots]

99% of the time 32-bit float would be an utter waste of space and time. The only scenario I can think of that truly warrants it is if your work will be the basis for some HDR lighting in a 3d package, or other specialty film-fx related need. I’m sure there are other uses, but it’s one of those “if you have to ask” things – if you needed it, you would know! 🙂

[Paul Nevison]

thanks

that’s what i thought….the difference between 8bit and 16bit I imagine would be better handling of banding issues?

Macpro 2x 3Ghz
4.0G RAM
OS 10.4.10
QT 7.2
FCP 6.0.1
AE CS3
BMD Decklink HD Extreme 6.6
PAL Land

[Darby Edelen]

Certain effects are much easier to accomplish realistically in 32bpc, especially with linear blending enabled. 32bpc obviously allows for a much greater range of values, so great that your display can’t even display the subtle variations between the values! So what is the point of all of these values?

In 8bpc and 16bpc all of the code values are mapped from 0 (black) to 1 (white) for compositing. In 32bpc the values are not limited by this range. If you try to push values above 1.0 in 8bpc/16bpc then the values get clipped at 1.0 (white), but in 32bpc the values are not clipped and are retained for compositing.

This can be valuable for emulating photographic processes. One of the most valuable uses for 32bpc is in compositing for film, which has a higher dynamic range than 8bpc/16bpc will allow.

In the end, most of the values generated in 32bpc will not be visible in your output (unless you are outputting to log Cineon/DPX for film).

Generally speaking there isn’t much of a use for 32bpc unless you are working on film or want to benefit from the unbounded range of values (useful for generating awesome light effects).

Darby Edelen
Designer
Left Coast Digital
Santa Cruz, CA

[Brendan Coots]

Yes, 16-bit can help reduce (or eliminate) banding issues and generally produce higher quality results. Even though you will ultimately be rendering out to an 8-bit codec, working in the 16-bit color space truly has advantages that are easier to justify than the jump up to 32-bit.

Darby’s explanation re: 32-bit is much better than my own. I agree that certain effects look more photographic in 32-bit – for example, when done correctly gaussian blurs transmit the pixel values behind them much more accurately and treat light values more like a camera might. Still not worth it though unless you are working with film.

[Sam Moulton]

almost all of my projects are 32 bit because everything that has a motion blur or a glow or moves works better, looks better, is just, well, better. CC is easier, blends are easier, compositing is easier. . .

The project size isn’t bigger, and I find that, in many cases, the render time doesn’t go up that much when compared with 16 bit projects. I do not render anything to formats that support float. Most are either to a 10 bit codec (black magic) or lossless animation, so my rendered files aren’t any different. I’d dig out a few of the float tutorials here or at total training and give it a go. Once you get used to the workflow and learn how to use the tools to integrate non float plug-ins, you’ll have a hard time going back.

[Darby Edelen]

[sam.mltn] “almost all of my projects are 32 bit because everything that has a motion blur or a glow or moves works better, looks better, is just, well, better.”

This can be true if you know what you’re doing. However, if standard footage or images (non HDR) are used in a project it can limit the dynamic range. In addition, certain blending modes do not work at all as expected in 32bpc.

In my experience, linear blending improves blurs and glows more than strictly 32bpc. However, when using linear blending you need to be working in at least 16bpc, and it works even better with 32bpc.

Darby Edelen
Designer
Left Coast Digital
Santa Cruz, CA

[Todd Kopriva]

Others have said this pretty well already, but allow me reiterate the statement that 32bpc is much better for lighting effects. If you work in 32bpc, you get the HDR advantage of overbrights. You do NOT need to use HDR footage to use HDR in your composition. Many effects can productively create overbrights if you let them.

The best example is a glow followed by a blur.

In 16bpc, you start with a light at brightness 1.0, the glow can’t increase the value above 1.0, and so the blur effect’s tendency to darken brings even the center of a light down below 1.0 quite quickly.

In 32bpc, you start with a light at brightness 1.0, the glow increases to (say) 2.0, and therefore the blur’s tendency to darken doesn’t bring the center of the light down under 1.0. The result, a light with a center that stays bright, even when blurred. Huge, huge difference.

I included a pair of images for comparison in the “Set the color depth” section of After Effects CS3 Help.

Of course, I wonder why I even bother sometimes. [*grumble, grumble*]