How Hard Is It To Create A Broken LUT?

Patrick, I too value the information gathered from generated ‘charts’. A thought is that even though an image looks alright, upon closer scrutiny one might actually find banding or some other artifact. It makes sense to QC a LUT with a chart sensitive enough to show where to look for possible issues. And, of course, there is always the question of just how much might one get away with in their film. But then I seem to be one of those who value nuances which may or may not at first be noticed. To add to your beta15.3 tips at the end: If one selects the picture viewer in the LUTs pane on the color page and scrubs over the various LUTs they will be temporarily applied and seen live in the viewer making for a quick way to change through several LUTs (or stills from the gallery.)

I’m no expert, but I’d say the reason of the inconsistencies in LUTs come because “commercial LUTs” are created by sampling chips: the various inconsistencies from one chip to another (both from developing, they use to be film-based, or from the capture of this samples) make the response non-linear. This subtle inconsistencies are streched when we play with the image -any LUT will require grading before and after it IMO.
Our grading tools affect the image “overall”. Lift/Gamma/Gain’s effect is very soft, affecting a wide tonal range: Artifacts are only added when playing too hard with Log controls, or by drawing curves with flat/too steep shapes: I think then artifacts will rise.
Old professional LUTs by Autodesk and even the ones in Resolve proved pretty reliable to me, but others one can buy for 100€ often got me into trouble: I guess sampling the whole latitude in 17 or 33 steps/channel could be the reason…
Food for thought, thanks for trying hard!

What a fantastic little ‘workaround’! Thanks for sharing, Pat.

I generated a Lut with Color space transformation and a lot of banding and artefacts appeared in the chart. I debayered BMD RAW as linear gamma, then a transformation to gamma Log C, then I worked in log mode, and finally, a transformation to Gamma and Color Space to REC 709 (BMD v3 to 709). I enabled only the transformation from Log C Nodes up to the end, and generated a Lut. When I tested, I saw a a destroyed chart. Is Color Space transf not allowed to be baked in a Lut?

From my testing: The LUT only encompasses actions performed on the Color Page. If you’re doing the transforms either in the Project Settings or in an OpenFX plug-in – the LUT export ignores those operations.

How are you executing those transforms? Using what tools?

Jeffrey – I’ve taken to turning off the Live Preview if I’m comparing LUTs. There’s a momentary flash as the mouse moves moves from one LUT to the next that makes it impossible for my eyes to do a direct comparison. When testing, I prefer being more old-school and applying then versioning and stepping through the versions for nice clean comparisons.

Actually – I’ve taken to turning off Live Preview everywhere. I just don’t like having my Viewers taken over without me actively telling the software to do so.

OFX Color Space Transform Pre Node for Linear to Log C,and a Post Node for 709.

My understanding: Currently OpenFX nodes are ignored. I’ve not heard that Transforms operations are handled differently from other OpenFX nodes. But if you need this, test it with each new Resolve 15 update. It may be added at some point.