Overall, it’s a great product. Can we expect a new update or is it replaced by Sharpen AI (which I like less after trying the demo).
I have a few observations that may help (or not) for future development.
- After using my raw images directly in Topaz Studio/AI clear, I just didn’t like the Studio controls for image shadows/highglights and so on. However I reckon using AI clear directly on the raw file produces much better results than on an already processed file. Which leads to the second point:
- processed images in ACR needs to be left with color noise reduction to 0 to offer the best denoise result. Maybe a set of training on images that have had already been color denoised? The results on such file is atrocious, leaving smudges of fat noise. Not a biggy anyway I saved defaults in ACR with no color denoise. But if AI Clear could have been trained on a set that accounts for color denoise, then it would work much better also on out of camera JPEGs in addition to quickly processed ACR/lightroom raws.
Here is an example, cropped 100%. Not sure which one will show first but the one with plenty of noise has received color noise reduction (default setting at 10) in ACR and the other has not.
- better and finer rendition of small details with medium and high denoise. It tends to fatten small details in fur/feathers and the likes and does not look very good even after a 50% downscaling. It looks near perfect in low denoise amount on files with reasonable noise.
- Less color shifting (if possible). The denoised result often shifts color quite noticeably, usually toward greens.
- a higher than high level of sharpen, but I guess that’d step on Sharpen AI’s feet ;).
Keep it up with “AI” stuff, way to go.
IMO, working directly with the Raw is not TL’s strong suit for any of their apps using the libRaw engine. I am not sure why as my go-to developer is also based on libRaw and does a great job. I always start TL workflow with a Tiff processed upstream with something else. I am rarely disappointed with the AI-Clear result using that approach. Having said that, AI-DeNoise is also challenged by images with lots of color noise. Some have found that reducing the color noise first with DN-6 is a reasonable work-around at this time.
April 28, 2019 email reply to an AI Clear inquiry:
Thank you for contacting Topaz Labs Support.
Glad to hear you are enjoying DeNoise AI. We are no longer selling AI Clear as it’s been replaced by DeNoise AI.
Please let me know if there is anything else I can help you with.
I doubt if there will be any upgrade\updates to AI Clear.
It’s now included with the new version of Denoise AI
I have a completely different experience where raw files without any type of color noise reduction are coming out much better, that’s weird.
wow there, I didn’t know that. Is there an upgrade path for existing AI Clear users?
As pointed out by pmeach and Topaz reply, DeNoise AI is the path.
Well this Denoise AI is great. I’ve been expecting an AI Clear update for some time, but thought that it was maybe still young enough to not receive one just yet
Thank you everyone for the pointer… I guess I missed the mail informing me about that, couldn’t find any.
Anyway, Denoise AI is quite perfect and has everything I would have wished for AI Clear to have without daring to ask: Weight sliders instead of hard presets and DNG output. Details fattening is now much better controlled or simply absent and being able to do the actual post processing on a dng after denoising the raw means a lot to me.
Thank you Topaz people, much appreciated!
Here’s my first picture processed with Denoise AI, the AI clear one was ok but the feathers details looked too fat.