What they're doing has been described in multiple academic papers but for one example: That's why it takes up 2 GB (!) of space for installation, we're largely paying for a large database of CV training data that has been trained using images and blur intrinsics. While the term "AI" is heavily abused by marketers, the product we're discussing here does use Computer Vision, Deep Learning models to de-blur. All sharpening is, is increasing the contrast on the edges of shapes, tonality, and colors. If you don't want or need that level of automation, it's perfectly OK to skip it! It does what other sharpening tools do! It just automates it. I use DeNoise AI on JPEGs, and keep it in reserve to use after DeepPRIME on the noisiest raw images. I nearly always use the motion blur mode. Yes, that's how I use them: every raw image is processed using DeepPRIME, and a minority of images are further sharpened using Sharpen AI. But it is still a more niche product ultimately, "AI"-based de-noise by contrast should be a staple since it is impactful on near every image. The only reason this product matters is that when it works it is magic, like "save an image" magic. There's a lot of images Sharpen AI does nearly nothing to, whereas a de-noiser is useful for almost every image but especially high ISO/long exposure (be it Topaz' de-noiser product or DeepPrime). De-Noise will take care of higher ISO problems/noise, whereas Sharpen AI is better at correcting a handful of optical abnormalities (like subtle motion blur and focus drift). Keep in mind their De-Noise and Sharpen are different products. Maybe i'm expecting too much of the software and not enough of myself. That said i'm trying to clean up 50mm long birds at ISO 5000 in the dark. I don't care if it takes 15 minutes to generate all previews, because I can be doing other work, but I do care an awful lot if I have to click every image to generate, and the previews are forgotten when I go back to an image. The program has groundbreaking tech, but you can tell it wasn't created by someone who has done volume photo finishing work, as working at volume is a complete non-thought. The closest you can get is "Zoom to Fit" and "Side by Side View" then let the preview regenerate (but this is largely flawed as the program has no concept of pre-generating/bulk generating the previews, so you cannot set it like you want then quickly grep through all your photos, it is this slow kludge while previews generate).Įven after purchase I just leave on auto, save all, then review the results via the JPG output to avoid the obnoxious "generating preview" every new image click (because doing preview gen for the next image in the background pre-emptively would make far too much sense). I get that they don't want people using it for free during the trial, but it made it much harder to evaluate. That bit of the Ai is perfect every time.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |