Darktable is great, but notably, it doesn't have any neural network-based denoising, even though that's now standard in Lightroom, Capture One, and other apps. Darktable only has rather outdated wavelet and non-local means denoising. So a photo that would be perfectly fine at ISO 6400 in other apps will still look grainy, or worse, splotchy in Darktable.
To give DarkTable credit, neural-network-based denoising will be in the next major release (5.6).
And even without neural networks, DarkTable denoising is better than open-source competitors, due to the database of camera sensor noise shipped with it. For each supported camera and ISO setting, it contains the measured values of Poissonian and Gaussian components of the sensor noise, so proper denoising becomes a one-click operation. That's as opposed to the much more complicated "drag the luminance and chrominance noise sliders until the noise disappears, then drag two more sliders to recover detail" workflow found, e.g., in ART.
Darktable has a "neural restore" algorithm [0] in the development version (intended for midsummer release). Note:
- It appears to be an out-of-band pre-processing stage (run the image through denoise to produce an intermediary TIFF), unlike most other parts of the program.
- All AI features are gated behind compile-time flags which default to off.
[0] https://github.com/darktable-org/darktable/pull/20523