Which bit depth does DaVinci Resolve grade in?

[Bouke Vahl]

[Juan Juig] “that affects only the preview quality.”

Not entirely true. On Mac, this also affects mixdown quality. (Not codec, but what is inside.)

Bouke
http://www.videotoolshed.com

[Marc Wielage]

[Bouke Vahl] “Try to totallly crash a shot, (crush all blacks, flatten out all hightlights). Then add a node, and bring it back. Normally, this would be impossible, but in Resolve this is no problem at all.”
That is not exactly true. It is possible to create a “destructive” correction early on in the process that can’t quite be undone later on in the node tree. So there are limits, even with 32-bit float.

[Bouke Vahl]

Ok, put your money where your mouth is.
Do show me a real world example that indeed will result in loss.

Bouke
http://www.videotoolshed.com

[Glenn Sakatch]

True, some plugins in Avid will tell you to switch to 10 bit before using them for best results in the analyze mode.

Adding a LUT at certain locations in the node tree I believe will cause destruction of the 32 bit float theory.

Glenn

[Marc Wielage]

[Glenn Sakatch] “Adding a LUT at certain locations in the node tree I believe will cause destruction of the 32 bit float theory.”
So will hard limiting or extreme keys. Do that, and you can’t get the information back. Lopsided crushed blacks (say -20 on Red, -10 on Green, -5 on Blue) is hard to normalize after that. Really screwing up Offset (Printer Lights) and the RGB Mixer can also wreak havoc on subsequent nodes. There are limits even to what 32-bit processing can let you recover.

I’ve warned students before that the principle of gain staging in audio also applies to video: if you badly screw up the signal early in the chain, it’s harder to undo that later on in the node tree. In particular, extreme gamma changes can create noise issues under some conditions.

[Sergio Parrella]

So does this mean that I don’t need to transcode my 8-bit 4:2:0 h.264 file to DNX/ProRes/Cineform to get Resolve to process it in a greater bit depth?

[Michael Gissing]

[Sergio Parrella] ” So does this mean that I don’t need to transcode my 8-bit 4:2:0 h.264 file to DNX/ProRes/Cineform to get Resolve to process it in a greater bit depth?”

Yes. Resolve does colour processing in 32 bit float regardless of the source files. You should output in at least a 10 bit codec to take advantage of the range realised in the grade.