The Argument Against AI Upscaling for Photography
Are we heading in the wrong direction? What's a better option?
B
ack in 2019 TopazLabs released a trio of post-processing apps for photographers that leveraged machine learning. They used a system trained on millions of photographs to cleverly remove noise, to sharpen an image and to upscale it.
The software was revolutionary, highly effective and reasonably priced. Other companies recognised the potential and incorporated similar features into their apps – and the rest is history. Most developers now include A.I. features in their post-processing apps and some (such as Luminar Neo) base the app’s entire identity and brand on its A.I. toolset.
But after those initial breakthrough years, the quality of some A.I. tools has actually declined – and the worst affected niche seems to be in upscaling where an over-emphasis on generative methods has many photographers reverting to old school methods.
What’s gone wrong? What should be done to address the situation and what’s the best current upscale tool for photographers? I have some suggestions.
The rest of this one’s for paid subscribers.
Already a paid subscriber or a Patron? Enter the members’ password from your welcome email.
Not a member? Subscribe on Substack or leave a one-off tip.
Also published in RAW & Unfiltered on Substack.
