Last year I shared a tutorial about an option Lightroom which can shrink your RAW files by 90% or more. In this tutorial, I want to share a much simpler workflow to compress the images as well as updates on what I’ve learned after using lossy compression for over a year.
This is an incredibly powerful tool to help get the most out of a laptop with a small drive, save money on external storage, etc. After working with lossy compression for a year, I feel very confident using it on a wide range of images. I just compressed multiple years of images, which resulting in freeing up 3 TB of storage on my RAID drive. That has helped me personally avoid probably $1000 worth of hardware costs (as I would soon need to replace six hard drives and my external SSD backup without freeing up this space). Most of you won’t be in that situation, but you may well offset much or all of the cost of your Adobe subscriptions by minimizing the cost of laptops, external drives, and backups.
In my original tutorial, I shared a workflow to use the export dialog to create the files. That is still the only option if you wish to create lower resolution DNG files (and can shrink the images by up to 98%, though obviously this degrades image quality when you change the resolution). However, if your goal is simply to save space with an image which is visually lossless (even for making large prints), there is a simpler option I had previously overlooked.
CAUTION: This is provided for your education only – incorrect use of this information (or potential software bugs) may result in loss of data. For example, failure to filter to raw files could flatten TIF or PSB files. Bulk conversion of images increases the risks. Do limited testing with duplicate images to understand these options and results before you do try any mass conversion of your images. You are responsible for your own actions.
To replace your existing RAW / DNG with a “lossy” (visually lossless) DNG, this is the simplest workflow:
- Filter for valid images
- Only real images, not virtual copies (as LR has a bug that will show an error, where it should simply ignore it as it is not a real file)
- Ideally, filter to file type = “Raw” or “Digital Negative / Lossless” (you should never convert layered files like TIF or finished exports like JPG, though the setting below should protect you even if you don’t filter by image type).
- You may optionally filter for images with zero stars if you have any concern about converting your best images. You’d get most of the compression benefit without altering your best images.
- If you have used
- Go to Library > Convert Photo to DNG and select the following options:
- “Only convert RAW files” must be checked (you do not want to flatten TIFs, etc)
- “Delete originals after successful conversion” should be checked (otherwise you won’t save space, or would have extra cleanup work)
- compatibility = “Camera Raw 16.0 and later” (this uses the new high quality JXL compression, older versions are not as compact nor as high quality)
- JPG preview = “medium size”
- “Embed Fast Load Data” should be checked
- “Use Lossy Compression” must be checked to get the size reduction.
- Do NOT check “Embed Original Raw File” or you will not save any space.
Note that once you have set these settings in that dialog, they are sticky and will be the default going forward. You can easily just search for “DNG” in the help menu to find the command if you forget where to find it and then just run with your previous settings.
Also note that if you have used LR’s merge to HDR / pano, the output is already a lossy DNG but will be compressed again if you select it (even when choosing “only convert RAW files”). The new result is nearly the same size and visually not degraded – but it would still be best to filter file types as noted above as this would avoid double compression of a lossy DNG.
Creating lossy DNG during import:
It would be ideal if Adobe would offer this an a simple check option during import. This would avoid the risk of picking the wrong settings, avoid re-compressing images merged to HDR / pano, make for a simpler workflow, save time, and enable users with very limited space to import more images (as you would need about 11x more free working space to import normally and compress than if you just imported directly as a compressed image).
Please vote for an option to import visually lossless DNG if you would like to have such a simple option (if you add any comment, please type your own thoughts – copy / paste never sounds sincere). Please note that the proposal is to create an option and have it off by default. I fully appreciate that most people will not feel comfortable using lossy (certainly if you have not spent significant time with it to see that the results are excellent) – so the proposal is to leave the default exactly as it is (ie, always lossless unless you consciously choose another option to save space).
Lossy DNG is a great choice if:
- You would benefit from significantly smaller files to save money on hard drives, work with limited storage on a laptop etc.
- And the caveats below do not affect your images (ie when the lossy file would not cause a visible loss of quality). Most images will show no visible loss of quality even for large prints from lossy files.
- You do not expect to enlarge your image more than 4x native resolution (I would have used this on all the weddings I used to photograph – would have probably saved 50-100 GB per wedding) or use exotic de-mosaicing options like DXO DeepPRIME.
- You are archiving old work (especially high volume events such as sports or a wedding).
You should avoid lossy DNG if:
- You need to use the files with versions of LR or ACR which are more than 2 years old (this is unlikely as you cannot install them anymore and any current subscriber can install the latest).
- If you use advanced de-mosaicing or wish to protect for the option in the future.
- De-mosaicing is the process of converting the sensor inputs (usually a Bayer pattern: 1 red, 2 greens, and a blue) into a regular RGB pixel. A lossy DNG has already done this conversion, which results in excellent quality using the best conversion available today – but theoretically future conversion tools may do even better (for pixel-level details such as color / luminance noise reduction, moire, and capture sharpening).
- This includes tools like DXO DeepPrime which only support the mosaic data.
- Avoid compressing high ISO images before you use AI Denoise.
- While Adobe now supports it even on lossy files, there may be color error. For example, in the deepest green shadows of trees shot at ISO 6400 in a night shot, I see patches where the trees show magenta.
- However, if I run AI Denoise before converting to lossy DNG, the lossy DNG is very good and highly consistent with using AI Denoise on the original RAW.
- In one example test image, the original NEF was 59.1MB, the lossy was 26.8 MB if denoise was applied before converting to lossy, and the lossy was 21.3 MB if lossy conversion was done before any AI denoise. So the lossy DNG is different if AI denoise active at the time the lossy DNG was created (though the amount had no impact, anything from 1-100 gave the same result and could be changed after lossy conversion).
- Avoid compressing if you use Adobe’s super resolution.
- There are artifacts when using super resolution on the lossy DNG which would be visible in a large print.
- If you turn on super resolution before converting to lossy, there is no concern. However, there is no compelling size advantage – the lossy DNG is huge (200 MB if lossy is created with super resolution active vs 21 MB if ). Applying super resolution to the NEF results in 238MB of data (as a 179MB .acr file is created).
- However, I see no such issue when using Topaz Gigapixel on the lossy DNG. The results are very clean. When comparing a 46MP image, I would be able to print at 100″ x 67″ (300 dpi) and see no difference. At double that size (200 x 134″), I see differences due to AI artifact, but I would not consider either version superior, just different.
- Bottom line: lossy DNG is fine for massive prints if you use Topaz Gigapixel, but I would skip lossy if you rely on Adobe’s super resolution (of course, it may improve in time to work better with the lossy data like Topaz does).
An alternative option to create a compressed image and retain the mosaic data:
Stan N (per comments below) contacted me to suggest another way to create compressed DNG files and retain the mosaic data.
- “/Applications/Adobe DNG Converter.app/Contents/MacOS/Adobe DNG Converter” -lossyMosaicJXL -mp -fl -d {{destinationFolder}} {{sourceImage}}
- Where you should leave a space after “-d” and then drag a destination folder from Finder to Terminal, add a space, then drag and drop your source image(s).
- -lossyMosaic creates the desired format.
- -fl keeps the “fast load” data.
- -mp allows it to process multiple files in parallel (ie much faster)
- -d indicates the output folder
This may offer additional value for you, but you should test carefully. My ISO 6400 image shows some changes in very dark regions. And while the output works fine in LrC, I was unable to benefit using DXO DeepPRIME (v5) as they apparently do not support this encoding. It will open as a very pink image and output a mangled result which is unusable. Hopefully support for this lossy yet mosaic format may grow in the future to offer the best of both worlds (I sent a bug report to DXO support).