• 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5

Workflow if using new Lightroom AI Noise Reduction?

Offline
#1 willsnow
Hi Folks

In using Lightroom (it's been a few years since I last used it!) I noticed the newly launched AI Noise Reduction which works really well, especially with drone DNGs.

However I saw that it creates a new DNG file ... so would this mean if we wanted to use it with LRT ... we would first have to run our source through this, get those new DNGs, and then open THOSE NR'd DNGs into LRT, and continue the same workflow?

Because if not, we would break the link with LRT because a new set of DNGs would be created mid workflow?

Thanks in advance for any thoughts

WSmile
Offline
#2 Gunther
Yes, you would need to do the AI noise reduction and conversion to DNG first, then separate the new DNGs from the unprocessed images and do the LRTimelapse Workflow on the processed DNGs.

Some warnings however:
- Doing the AI Denoise on a timelapse sequence will take a looooong time.
- Since this is a non linear process, you might get different editing from image to image, it's not predictable.
- In my experience you can get excellent denoised results by using a combination of the "normal" luminance/color noise filter in LR and especially the "Motion Blur" when rendering in LRTimelapse. By blending the images in Motion Blur you get the noise away very well.
- Personally currently I see no point in using the AI Denoise for timelapse.
Subscribe to: LRTimelapse Newsletter, Youtube Channel, Instagram, Facebook.
Offline
#3 Mhelit
What motion blur is good I usually pick 5
Offline
#4 Gunther
It depends on your sequence. The faster the movements are, the lower the Motion blur should be. Best render different versions with different MB settings then you will see the difference.
Subscribe to: LRTimelapse Newsletter, Youtube Channel, Instagram, Facebook.
Offline
#5 adilfariz
Hi Gunther,

I don't understand your step here. " Conversion to DNG first, then separate the new DNGs from the unprocessed images and do the LRTimelapse Workflow on the processed DNGs. "

After I denoise, it will create a DNG file. What I need to do on the step workflow?
Offline
#6 Gunther
You need to denoise the whole sequence in Lightroom. This process will create DNG files for each image.
Then you need to remove the original files from the folder, before you do the regular LRTimelapse workflow on the folder with the denoised DNG.

To remove the original files, for example you can load the folder into LRTimelapse - you will then get a message that there are different file types in the folder and that LRT will select all images with the least occurence, which will either be the DNG or the original images. In either way, now you can right click on one of the selected images and do "Create new folder from selection" to move those files into a new folder and separate them.
After this, load the folder with the DNGs into LRTimelapse and process as usual.

Before starting the workflow, you should remove the old folder from Lightroom (because in that folder are still references to both file types). You will add the folder again during the LRT workflow.

This is more complicated to describe than to do :-)
Subscribe to: LRTimelapse Newsletter, Youtube Channel, Instagram, Facebook.
Offline
#7 willsnow
Many thanks Gunter for your detailed reply, apologies for not replying sooner I hadn't checked the forums since posting.

I'll stick with regular denoise for now - it seems good enough Smile
Offline
#8 adilfariz
meaning, if I have 300 shots, i need to denoise 300 shots? before this without denoise, we only edit the keyframes only right?
Offline
#9 Gunther
Of course, bacause that AI denoise is not a "normal edit" in Lightroom, it's an image conversion.

Personally I find it tedious and too time consuming for the little bit in quality improvement (if at all) that it delivers on timelapse, as I wrote in my first post. But try it out for yourself and decide if that's the route you want to go.
Subscribe to: LRTimelapse Newsletter, Youtube Channel, Instagram, Facebook.

...also check out: