Hello Gunther and the community !

I just shot a "day to night to stars" sequence with a manual lens. So I've input the aperture value when I was asked for. After that, when I activated the holy grail wizard, it gave me a strange curve, which is huge and with enormous jumps (see screenshot #1). Usually, my holy grail curve is (almost) the exact mirror of my luminance curve, but this time it was really weird. I managed to get it back to a decent curve (though it wasn't perfect) by tweaking the two sliders.

But I also noticed that at the very beginning of my sequence, my luminance curve remained pretty even, but the HG wizards detected some jumps (but there wasn't any) that it has tried to compensate (screenshot #2). You can see the luminance values are almost identical but the HG applied two compensations very different. It resulted in a visual preview where the HG has literally CREATED jumps where there weren't (screenshot #3)

I tried to review everything to see if I did something wrong but didn't find anything. So I have 2 questions :

- why did the HG wizard behave like this, creating this curve that is crashing against the upper limit of the screen ??

- why did it "find" jumps to compensate where there wasn't any ? Did I mess something ??

thank you for your help !

Generally I don't think this is a bug - I will explain how this could happen and what to do.
But it might be a good idea to carefully check the table - holy grail adjustments will only happen, where ISO/Aperture/Exposure have been changed. And they should compensate for the extent of the changes.
To detect the changes, LRTimelapse will rely on the Exif Data from the camera. And it will expect the Changes in Shutter/Aperture/Exposure happen in steps of 1/3 stops. This should be default on all cameras, but some cameras offer an option to work in 1/2 stop steps - this is not supported by the HG-Wizard. Please double check.

Ok now for the workflow.

If you do really long transitions which span a big difference in luminance and that are not very even like in your case (luminance curve goes up and down and the adjustments have different spans) this could happen.
LRTimelapse tries the best to fit a compensation curve, but it's normal that you need to twist it (that's why you have the sliders).
There might even be cases, where it's impossible to fit a compensation - the reason is that the compensations while shooting were too uneven. Using a ramping tool like qDslrDashboard is always a very good idea in order to get better results.

Ok, but now on how to save your sequence: first thing I would do is set a reference area (drag a rectangle to an area of the sequence which has only sky) - don't cover the sun with the reference. This might already give you a better luminance curve. Now you can reapply the HG Wizard. See if that helps.

thank you very much for the answer. In fact I always change my settings by a 1/3stop, but in this case for example, I had the sun in my shot and an overexposed area that I had to let disappear before raising my shutter speed again. that's why my luminance curve sometimes goes down long before I correct it. But beside one or 2 times maybe, I always does that with 1/3stop increments.

Im gonna try the reference area tip !

for your information, in the meantime, I removed a lot of the orange rectangles (holy grail keyframes) into the right panel, and the curve went back to normal after recalculating the holy grail wizard without all these holy grail keyframes. I don't get why...

Did this triangles appear on places where there were no changes in aperture/iso/exposure? Normally they don't. please check in the table. All information ist there.

But yes, it's also a way to solve such problems. After all the visual deflicker will be able to do compensate for those jumps also with a couple of passes.

Yeah they were properly disposed. That’s why I find it weird... anyway I recovered a good curve and will try with the reference area in the sky. Thank you Gunther.