This post was last modified: 2017-10-01, 23:19 by ridindave. Edit Reason: add screencaps

Hi Gunther,

I've recently upgraded to LRT4, while pretty happy with LRT3. Even after reading your instructions many times, I'm still experiencing weird results, and have numerous questions. I'll try to break things into simple questions...

- when using HG or deflicker, I understand classic exposure setting is not used anymore. Instead I've found it's somewhere hidden inside gradients. Is it normal we can't access no more to these values from within LR ? I see 4 preset gradients but none of them handle the HG or deflicker exposition variable.

- still using HG, there is a column named "HG lum leveling" just next to the classic exposure column. What exactly are the numbers I read here ? Obviously, there are not exposure values, because where I should get 1/3 exposure difference because of a 1/3 exposure variation when shooting, I go for ex. from -0.032 to -0.015. which is very far from 1/3. Still, it seems to work, but those numbers are total mystery to me...

- I'm working on a stereoscopic shot, where I've badly set my cameras. One eye is full raw, the other one is sRaw2 (half definition raw), instead of both sRaw2.
Both eyes were shot stepping exposure at sunset, so I need HG feature on both. Runs smooth on full raw, but not with sraw2. Seems that HG works ok, but when using LRT visual previews I have unexpected crazyness from time to time. What happens here ? Is this because of Canon sraw2 ? I'll try to convert to DNG first but I've really never had to do it, since workflows are usually long and complex enough for trying avoiding this extra step. LRT3 was always fine with sraw2, btw...

Here are 2 screencaps, fullframe first, then sraw2. Same settings applied of course on both :

https://i.imgur.com/VPTnaP0.jpg
https://i.imgur.com/xEe3w8D.jpg


- About visual preview : some days ago I've noticed hidden folder within the raw folder I was working on, containing visual preview thumbs. Now I don't see it anymore with other sequences, and when I switch from one folder to the other I have to recompute visual preview each time to get it back displayed. Aren't those supposed to be cached ? Is this something we can set somewhere ?

- I also went back into a really old sequence, that used to be tweaked with previous LRT version. I've trashed any previous xmp files, renamed the raw files (had to do it for reason). Where I should have started with something LRT should think is new to him, I was still left with previous settings, such as crop window and some other "ghost remainings". How is this possible since name changed, xmp were trashed and erased, etc ? I remember previous versions used stubs. Does LRT4 also uses them ? If yes where are they stored ? Can't find any!

Thanks a lot for your precious time, hope I don't waste it too much with my questions...

cheeers,

David

- when using HG or deflicker, I understand classic exposure setting is not used anymore. Instead I've found it's somewhere hidden inside gradients. Is it normal we can't access no more to these values from within LR ? I see 4 preset gradients but none of them handle the HG or deflicker exposition variable.

Yes, you can't access those values anymore and you don't need to. Just let LRTimelapse do it's magic.. :-) One important thing related to this is to not remove any from the predefined gradients and not add any new gradient filters since this would break holy grail and deflicker.

- still using HG, there is a column named "HG lum leveling" just next to the classic exposure column. What exactly are the numbers I read here ? Obviously, there are not exposure values, because where I should get 1/3 exposure difference because of a 1/3 exposure variation when shooting, I go for ex. from -0.032 to -0.015. which is very far from 1/3. Still, it seems to work, but those numbers are total mystery to me...

It's stops/4 I think, but honestly, it doesn't matter. Just ignore the values, let LRT deal with them. Check the curves, that's what matters.

- I'm working on a stereoscopic shot, where I've badly set my cameras. One eye is full raw, the other one is sRaw2 (half definition raw), instead of both sRaw2.

Both eyes were shot stepping exposure at sunset, so I need HG feature on both. Runs smooth on full raw, but not with sraw2. Seems that HG works ok, but when using LRT visual previews I have unexpected crazyness from time to time. What happens here ? Is this because of Canon sraw2 ? I'll try to convert to DNG first but I've really never had to do it, since workflows are usually long and complex enough for trying avoiding this extra step. LRT3 was always fine with sraw2, btw...

Make sure to have both in different folders. Make sure too to have the latest Adobe DNG Converter installed. If development of any RAW files goes wrong, it's the DNG converter, LRTimelapse itself doesn't develop Raw files.

- About visual preview : some days ago I've noticed hidden folder within the raw folder I was working on, containing visual preview thumbs. Now I don't see it anymore with other sequences, and when I switch from one folder to the other I have to recompute visual preview each time to get it back displayed. Aren't those supposed to be cached ? Is this something we can set somewhere ?

All LRT previews (camera previews and visual previews) get stored in a hidden folder called ".lrt" inside each sequence folder. Previews from LRT3 will get migrated automatically to that new location, if possible. Of course any previews get cached there. The visual previews will only get recreated, if the metadata changes. If you set your Lightroom to automatically write changes to metadata for example (not recommended at all) you might get this effect. Make sure, that the XMP files are only touched when you manually hit save in Lightroom.

- I also went back into a really old sequence, that used to be tweaked with previous LRT version. I've trashed any previous xmp files, renamed the raw files (had to do it for reason). Where I should have started with something LRT should think is new to him, I was still left with previous settings, such as crop window and some other "ghost remainings". How is this possible since name changed, xmp were trashed and erased, etc ? I remember previous versions used stubs. Does LRT4 also uses them ? If yes where are they stored ? Can't find any!

If you right click on a folder in the LRT folder tree, you will find the new option "Clear all LRTimelapse edits" - do that to get a really clean start.

Hope I could help clarifiying those things. Just practice a bit with LRT4 - I'm sure that you'll love it - it's so much easier and more comprehensive as LRT3 and delivers much better results in terms of deflicker also.

Thanks a lot Gunther for all your replies. I will practice and get used to it, no worry here!

About deflicker, you state it's better than previously, which is probably true, but I feel it's still missing some control.
About strength, usually the minimum I can set is already too much. I just want to correct aperture fluctuations and nothing more. And I do really find the least setting will correct more than I want. From this perspective, I really wonder what the stronger settings are for...
I also feel we should have a control over time. You probably had to choose some mean operation that runs across a number of frames. Is this a fixed number ? Or a fix % of total sequence frames ? I feel it behaves differently if my seq is 100frames or 10k frames. Am I right ?
Anyway, having some kinda 'width' control would be really helpful, as it would help adapt the deflicker to own needs.

Thanks again for your attention,

best,

David

Hi David,
deflicker is much better, since it now uses the developed previews (visual previews) instead of the camera previews. This takes the Lightroom edits into account. And you can add "refine" steps to get even better results by doing multiple passes.

The one and only slider you need now is "Smoothing". You can adjust very subtile, which are short term flicker effects and which are long term natural effects. I do not recommend to set the value too low. Additionally it's recommended to set a reference area for deflicker. Choose a part of the frame, where only flicker is visible and not changes in brightness that you'd like to preserve. Define the reference area (as in LRT3) just by clicking and dragging on the preview.

Well, what I'm trying to tell you is that smoothing at its lowest is really far too much (I mean it happens across a too large number of frames) when I use very long sequences (several thousands). What I suspect is that is may be scaled to shorter sequences but not the ones that are that long. If you want clearer statement, I can fill in with screen caps, and you'll agree that lowest setting is already taking too many frames into accounts, no matter what you may actually think... I understand I may have specific issue because of really long seqs, but even if you recommended not to go too low, it's way too high. What about the inner logic inside deflicker ? Is this something secret ? Or shall we know if strength is based on specific number of frames, or depends on seq length ?

Thanks again for your patience and attention,

Cheers,

Dave

Can you please send a screenshot of an example where you think that the lowest setting is too high. Deflicker of course needs some room for smoothing, otherwise it wouldn't make sense at all. The aim of deflicker is to generate smooth sequences that are visual appealing to the eye.

Hi Gunther,

On this specific timelapse, you'll see I'm running into trouble both because of the timespan being too wide , and correction strength, set too high.
It's a very specific timelapse, as it is a boat travelling. It's all but a fixed shot, and to get worse, boat goes under bridges many times. So my analysis zone being the sky, it's filled with dark from time to time when I pass under bridges...
I've decided dynamic zoning was not an option, as it is a stereoscopic shot. I have 2 cameras trigged in sync, so I should apply very same deflicker correction to both camera, given still that each camera/lens has its own inner flicker.

As needed deflicking here is only to correct iris inconsistencies, when closing to the f8 setting I went for, I would say it won't go at its maximum over 0.1 exposure, probably less. So what I would love to be able to do here is limiting maximum deflicker delta to 0.1 exposure. This is one example where strength can not be set the way I would need.

On another perspective, timespan considered here is also too wide at its minimum. You'll find below on second screenshot, that is a portion of about 300 frames within the same sequence (that is 13 000 frames total). See how large is the timespan taken into account to smooth the curve. Should I get the same result if my sequence would only be these 300 frames ? I'm not sure...

Did you get my points here ?

Whatever, thanks again for your time and attention!

best,

David


• first screen is the whole 13 000 frames
• second one is a closeup of 300 frames within it
• last screen is whole sequence with deflicker applied so we can see amplitude over time

https://i.imgur.com/7XwKMjB.jpg
https://i.imgur.com/xxcYP4J.jpg
https://i.imgur.com/WcUovHP.jpg

About the 300 frames close-up, I've answered myself to the question. Here is what it looks like when deflicker is applied to an only 300 frames sequence.
So your deflicker logic behaves differently overtime depending on the sequence length...


https://i.imgur.com/a9QGLlH.png

Yes, I agree, it's a very special case. Regarding the bridges, even if the resolution of smoothing was higher, it wouldn't be able to distinguish shortly between shading due to a bridge and flicker.
This is just a scenario that LRTimelapse is not made for.
Normally 13.000 images is nothing anyone would look at. It's a clip that plays over 7 minutes. Normally you would use clips of 10 seconds max then cut to the next. Everything else really gets boring quickly.
I'd recommend to cut out some interesting subsequences and work on those individually. You can easily split the sequence by selecting a subsequence in LRT, then right-click and "new folder from selection".
LRTimelapse is not designed for such long sequences. For a reason the maximum supported sequence size is told to be 10.000 in the LRTimelapse user license. https://lrtimelapse.com/eula

This post was last modified: 2017-10-05, 12:33 by ridindave.

Hi Gunther,

I know this is a long sequence, and don't expect you to think otherwise. But still, your deflicker acts differently based on sequence length, and I think this is some point you should think about for the next major upgrade.


On another hand, what has not to do with length is the ability (the lack of, here) to restrain deflicker under a certain strength, or exposure delta. I'm pretty confident that this would be a feature that many serious users would warmly welcome, don't you think ?

Thanks again,

D.