How to pick colors above 1.0

May be an ignorant question, but I can't find a way to pick colors above 1.0 as if picker clamps the values. Searching the manual, forum and google didn't give any info on this.

Hi Hendrik,

The picker will get values above 1.0 as long as the image under it is "float16" or "float 32" and, of course, it has bright areas that extrapolate this level.

For me this values above 1.0 appear mostly when I use the "Log to Lin" tool or in 3D renders set to "float16" or "float 32"

Ok, thanks, I think I got where the catch was.

As I understand, when an image is read as int8 or int16, that pixel format flows through all downstream nodes even if project, viewer and everything else is 32bit float, causing every node to clamp? For example cranking gain up in brightnesscontrast node does not produce values above 1.0 even when clip checkboxes are not checked..

Is there a clear logic that can be followed, for example how do nodes that have multiple inputs set output data format?

You can always force a loader to treat the image data as 32 bit float in the loader settings. This will help get the most out of your footage when doing extreme stuff with it.

I undestand that, although going above 1.0 is a frequent affair, not extreme,

But still, how does Fusion handle streams with mixed input formats? For example, if I add two inputs, does the base input set output format? How are mask values handled?

The image format is always inherited from the Background (gold) input. It's a good idea to set your Global Preferences to 16 bit float in the Frame Format section. That way, your footage should always be converted to float, no matter what the source was. I'll need to check on that when I am at home with a vanilla install. You may need to set a default on the Loader to have it choose the Default depth instead of Format.

Masks are usually int8 unless you deliberately change them.

You can change the defaults for a tool by setting it up the way you like, right-clicking the node and choosing Settings > Save Default.

Thank you for the info!

The confusing part is that global settings were all at 32bit float, so I expected all inputs to be converted to that. But Loader defaults should also be set properly.

Back at home with my vanilla install. The Loader definitely defaults to Format depth, so changing that is recommended.

One of the reasons that Nuke is slower than Fusion is because it always works in full float. There are many cases where a comp that might have float elements doesn't need to be entirely float. Int8 is always going to take up 1/4 the RAM which means less re-rendering, and it's usually faster than float too, sometimes by more than a factor of 4.