-
Notifications
You must be signed in to change notification settings - Fork 1.2k
Camera support
Every new camera means new things to explore for both camera manufacturers and software developers.
Open source software like darktable doesn't have direct connections with manufacturers to get raw format specs and SDK before camera releases (nor is that feasible w/o and NDA), so we have to depend on samples and data provided by users, and reverse engineering by developers.
Here's all the steps involved in getting new cameras to work with darktable (and for older cameras to get better).
The first step in getting a camera supported is providing raw image samples.
We are looking for shots that are:
- Well-lit, pattern-full, scenery, low ISO. For example, a daylight landscape
- NOT a color target
- Lens mounted on the camera, cap off
- Image in focus and properly exposed
- Landscape orientation
We are NOT looking for:
- color targets
- Series of images with different ISO, aperture, shutter, wb, lighting, or different lenses
- DNG files created with Adobe DNG Converter
- Photographs of people, for legal reasons.
If your camera produces several types of raws (like different bit depths, compression modes, aspect ratios/crops, etc.) please take shots using all those types of exactly the same content with same exposure params, settings etc.
Photos have to be released under the Creative Commons Zero (CC0) - public domain licence.
Upload your samples to https://raw.pixls.us/.
Also, check this thread periodically: https://discuss.pixls.us/t/raw-samples-wanted/5420?u=lebedevri
Once you've uploaded a full set of samples, please open a new Camera support request.
Please keep in mind that providing support for raws is a lengthy process, so before starting, check if somebody else hadn't already uploaded samples and filed appropriate issues. Check also closed issues because between the last darktable release and now the support might have already landed and so you'd be doing all the work unnecessarily.
In order to get your camera's white balance presets in darktable's white balance module, please follow the steps described in our white balance presets wiki
darktable has a great denoising tool in form of denoise (profiled)
module. In order to get the best possible results, a profile has to be created for every ISO setting of your camera.
Follow the How to create camera noise profiles for darktable tutorial and send us the resulting package by creating a noise profile issue on Github
Whenever you're shooting an image, there are always imperfections in lens design distorting images slightly (or considerably) that need to be corrected. darktable uses the lensfun
library for lens correction in the lens correction
module and doesn't have its own lens database. If your camera and lens combination is not recognized, try updating your lens database using:
lensfun-update-data
If that doesn't work, then maybe your camera/lens is too new and not supported yet. You can check lensfun lens coverage to see whether your lens is on the list.
If it isn't, follow the creating lens calibration data for lensfun tutorial.
NOTE: Starting from darktable 3.6, scene referred workflow (based on filmic rgb instead of basecurve) is the default one.
The purpose of the basecurve is to make the otherwise scene-referred linear (linear raw rgb) color look good on your output devices.
Nn darktable we try to match approximately how vendors process their images by guessing the basecurves visually/manually. If your camera is not supported yet or you don’t like the result or just want to play with it, there is also a way to measure the basecurve from a raw and JPEG pair.
To get the best results, use similar pattern as in the Noise profiles section for shooting a sample image.
Remember to disable all "enhancements" your camera offers and shoot using the camera default image style (usually called "standard").
And finally, the follow basecurves tutorial on the darktable website.
To get almost perfect OOC JPEG feeling from your raws automatically, it is required to create a color profile.
First you have to get yourself a good color chart as typical ColorCheckers usually don't have nearly enough colour patches to be worth working with. It's best to get an IT8 chart as a target. If you do have a ColorChecker or something similar, the results might still be acceptable.
Follow the Profiling a camera with darktable-chart tutorial.
darktable wiki is licensed under the Creative Commons BY-SA 4.0 terms.