Culling photos fast

The biggest pain point for me, and identified area of improvement, is how fast I can go from clicking a thumbnail, to viewing the full size image.

I am dealing with 24mp raw images. Running Intel 4690k, 8gb ram, current samsung ssd over sata.

Concrete example. I was shooting my friends family christmas pictures with two toddlers who don’t like to pose. So I was just shooting almost continuously to hope I got a good shot. Now I have 200 raw images to go through to find the 2-3 keepers. All of them look the same from the thumbnail. So I will be clicking and opening image after image…and it end up being torture…

Firstly, is there some setting or workflow that I am not aware of that can improve this?

Secondly, if there isn’t, I think there should be one added. For example, when I open a folder of RAW images, maybe as an option, it could demosaic (and other process) all of the images, and put all of the resulting image data into ram. Then clicking between the thumbnail and full size image would be blazing fast. This could take a lot of ram, but hey, ram is cheap, and if this feature existed I would updgrade to 64gb or 128gb tonight! Using this feature, you could load a folder, then go grab a coffee, then come back and then work super fast!

Or there could be a preload option, that does the same as above, but on a user selected subset of images.


RawTherapee has some awesome CPU optimizations, however, Intel says you have 4 cores and 4 threads from a processor released in Q2 2014. That is probably your main problem when it comes to speed.

I use geeqie to cull my images, it loads the thumbnail preview from the raw file quickly and I can usually determine if its a keeper from that alone.

If that doesn’t work for you, just send all your raw files to the batch queue, make them jpegs and look through those.

1 Like

Thanks for the tips! I will try them all out (except upgrading my cpu :stuck_out_tongue:).

I still think even with a fast CPU, the suggested feature would be beneficial.

Reading an image from a blazing m2 ssd, and demosaicing on an 18 core beast CPU maybe be fast…but reading bytes from RAM will always be way way way faster.

Another possible improvement would be to save the post demosaiced image as a file next to the raw image

Also look at Preferences > File Browser > Cache Options > Maximum Thumbnail Height

Edit: should’ve found this option first.

You could use the Inspector to find the keepers. It shows you the embeeded jpeg, which is much faster:


Thanks again…another great sounding suggestion!

The inspector was not ideal for achieving what I had hoped with it. Thanks for the suggestion anyways!

Unconstructive criticism is not welcome here. You need to be civil.


Sorry. Is there a way to have to have the inspector preview scaled down?

I use Geeqie for culling as well. If it’s a large shoot I’ll use Geeqie to mark the best images, then copy those to a different folder to work with in RawTherapee.

In shoots where the out-of-camera jpegs are good without the need to process in RT, but I shot only raw, I use Exiftool to extract the embedded jpegs from the raw files. Here are the commands I use. It’s very fast.

exiftool -b -JpgFromRaw -w .JPG -ext nef -r .

exiftool -tagsfromfile %d%f.NEF -r -ext JPG -overwrite_original -r .

1 Like

Not yet. But perhaps I find the time to port this functionality from ART to RT :wink:


May be heresy here but I use non open source software for culling. :slight_smile:

It was very cheap to buy, so I just move the images I like off the SD card to my hard drive with it and then work on those in RT.

  • I get Raw+Jpeg from the camera.
  • I have a script that move the Raw to a subdirectory
  • I cull the pictures using the full-size jpeg
  • I run another script to mark the Raw that have become orphan.
  • Likewise I sort the survivors into subject directories using the JPEG, and then I run a 3rd script to find the matching raw and move them to the appropriate subdirectories.
1 Like

Is it a bash script? Can you share?

Very basic scripts with hard-coded Canon naming conventions… No warranty, etc…

My set up is that I have the JPGs in a directory, and the matching CR2s in a subdirectory called cr2. When I sort by subjects, the directories to which I move the JPG files are usually siblings of the directory where they have been kept together.

  • subcr2: the one with which everything starts. Creates a cr2 subdirectory and moves all the CR2s to it.

subcr2 (265 Bytes)

  • orphancr2: marks as “orphans” all the CR2s without a matching JPG (by renaming them), in other words all the CR2s that match a JPG that didn’t survive the culling. The script used to be more ballsy and erase the CR2 right out, but a bug in the script made it erase CR2s I should have kept (they were still on the card, phew). So now the final erase is a manual step.

orphancr2 (321 Bytes)

  • getcr2: searches the sibling directories for CR2s that match the argument JPGs, and puts them in a cr2 subdirectory. I use it after sorting the JPGs to various directories to reconcile them with their matching CR2. Code has been improved to retrieve only CR2s that are about as recent as the JPGs (an early version made a mess after the image counter rolled over).

getcr2 (523 Bytes)

1 Like

Hi @Ap0c552, I’m on GNU/Linux, I strongly recommend using geeqie for culling, particularly if you have thousands of images from an event and you need to deliver them in a few days.

+1 on Geeqie. A bonus is it pairs the raw and JPEG files, and delete gets rid of both at once.

1 Like

I used Geeqie. I like it! Thanks!

I still wish rawtherapee was this fast…

I don’t really know what kind of magic are able to spell the developers, but following your concrete example and a bit of maths from the Rawpedia documentation, your 200 images could need around 100 GBytes (Gigabytes) to store them on memory, without taking into account all the background processes and applications running in your system.

If you’re happy with Geeqie, keep using it. But keep in mind that it works with the embedded jpeg present in raw files, and that’s why its much faster than RT.

1 Like

Ya it would definitely be an extreme feature that could use some thinking. But hey if you want to spend like $35000+ on a PC, you could get the new Mac Pro with 1.5tb memory and load like 2500 raw in memory in rawtherapee :stuck_out_tongue:

But more reasonably, I think rawtherapee could improve how it deals with displaying the embedded jpg (Inspector). As it stands, I am quite disappointed with the inspector.