An easier way to work with log footage - Tutorial

[Juan Melara]

Hi all, I've just a launched a new series of Davinci Resolve tutorials. First up is a couple of techniques to make grading log footage a bit easier.

I've already had a question about how these techniques are applicable to the IPP2 workflow. Basically the REDWideGamutRGB / Log3G10 combo is already a great starting point. But its still possible to transform it into Rec709 gamma and then back again. The technique I pretty much always use with IPP2 is to remain in REDWideGamutRGB / Log3G10 right through the entire node graph. Then if I'm grading under a film print emulation LUT I create a Color Space Transform node just before, to transform RWG into Rec709. Gamma stays as Log3G10.
If I wasn't grading under an FPE I would do a gamma transform from Log3G10 into Rec709 in the same CST node as well.

Hope this is helpful. Let me know if you have any questions.

Corresponding blog post: http://juanmelara.com.au/blog/an-easier-way-to-grade-log-footage

 

[muhammad]

Awesome

 

[Gunther Meisse]

Thought I finally had this all figured out, shit I am confused again.

 

[Ben Scott]

Thanks Juan

 

[Peter Moretti]

Juan, thank you very much. But I guess I just don't understand why the OFX conversion is better than the LUT. I wouldn't do a grade before the LUT, so I don't see how the clipping of info before the grade matters. Testing it out, it does appear that if you do a grade before the OFX plugin and clip the values, they *can* be recovered in subsequent nodes. So is that the main advantage of using the OFX plugin vs a LUT? Thanks much and thanks again for taking the time to put together this tutorial!

 

[Juan Melara]

The benefit of using the OFX plugin is in being able to transform into another colour space/gamma for specific operations and then back again, non-destructively. Add in the adjustable gamut mapping and luminance mapping to all the possible transform combinations and you have a plugin that can't really be matched by LUTs. At least not without custom generating a LUT for each possible scenario.

 

[Cameron K]

Beautiful done Juan, and thanks for the FS700 Slog 2 technique. A lot of us still have these great little cameras.

 

[Marcus Friedlander]

The benefit of using the OFX plugin is in being able to transform into another colour space/gamma for specific operations and then back again, non-destructively. Add in the adjustable gamut mapping and luminance mapping to all the possible transform combinations and you have a plugin that can't really be matched by LUTs. At least not without custom generating a LUT for each possible scenario.

Hey Juan!

Thank you for the tutorial! Seems like a great idea!

I do have a question about the above statement that I bolded. Do you mind elaborating on how this OFX transform is non-destructive? I was under the impression that all transforms were destructive in some way shape or form? And that the goal should be to arrive at the final image with the least amount of transforms done to the image as possible, for the cleanest possible final signal.

Am I misunderstanding something?

Thanks again!

 

[Juan Melara]

The transforms are non-destructive as they're mathematical formulas. There's no loss of quality transforming to and from as they're carried out in floating point space. Watch out for luminance mapping as that will clamp values and you'll lose information. The same goes for transforming in and out of Red Gamma 3 or4 as that clamps as well.

It's not the end of the world though, just be aware of what can happen. And try to place anything destructive towards the end of the node graph.

 

[Michael Tiemann]

The transforms are non-destructive as they're mathematical formulas. There's no loss of quality transforming to and from as they're carried out in floating point space. Watch out for luminance mapping as that will clamp values and you'll lose information. The same goes for transforming in and out of Red Gamma 3 or4 as that clamps as well.

It's not the end of the world though, just be aware of what can happen. And try to place anything destructive towards the end of the node graph.

I think a more mathematically precise way of saying it is that the transforms are invertible, meaning they have an Inverse function. You can do lots of things with floating point math that's not invertible (such as multiplying by zero). But if the LUT has the property that it's invertible, then the original can always be reconstruction by applying the inverse LUT.

 

[Marcus Friedlander]

Thank you both for your answers!

There's definitely a ton more reading I need to do on this subject!

Do either of you have suggestions on where I can start?

Thanks!