I am asking myself - looking at many of the contributions, some about one or two years ago, how much AI should be permitted to use without mentioning it? I know, almost all, what we do to get a “nice” and impressive image, is at least AI-supported or based on selected AI algorithms.
But I am getting concerned about the option to add details which cannot be there if you calculate the optical performance possibilities of a scope-camera-seeing-darkness combination.
Not that I am against using intelligent “sources” to manipulate images, but as in “normal” photography, I think, it should be mandatory that the creator mentions it in the technical details.
What do you think about this idea? Is it nonsense or a good one?
CS
Georg
How much AI in astrophotographs without mentioning it?
Well written Engaging
I think this is a hornets nest, but I completely agree. That is as long as we are referring to generative tools that add, change or manipulate anything, i.e. makes the appearance of something look different from what it really is in any way. I’m not a fan at all and don’t think it has a place in any kind of photography. That doesn’t mean that I am against it either, just that it isn’t photography anymore. It’s closer to synthetic imagery or “synography” even though it might be a blend in some way.
Thus I think it should be separated and treated as such. I don’t buy the “but when we process an image, it’s also altered” argument because the main thing is the origin of what the image represents. I.E. Real light captured off of real objects using physical lenses and sensors vs constructed by algorithms and generated by prompts.
I think it’s comparable to an athlete using performance drugs to enhance the body’s natural physiological limits. Generative AI is doing exactly the same by artificially enhancing images and pushing them beyond the physical limits of the optics, sensors and captured photons. Which in my opinion renders it worthless, but for those into synography or digital art may be the bee’s knees.
I don’t think non-generative (but AI powered) tools that make manual/tedious processes quicker/easier to do needs to be listed. I wouldn’t specifically write it if I use deconvolution or unsharp mask etc. Though I’m a strong believer that people should spend time to learn the processing software well enough to understand this for themselves.
Well written
Lately I’ve felt that I haven’t gotten as much enjoyment out of astrophotography processing as I used to and I think the AI tools are a big part of that. I enjoy AP because it’s challenging. So for the next project I think I’m going back to basics and abandoning all of the AI tools altogether.
Well written Respectful Engaging
Georg N. Nyman · Jun 21, 2026 at 12:29 PM
I am asking myself - looking at many of the contributions, some about one or two years ago, how much AI should be permitted to use without mentioning it? I know, almost all, what we do to get a “nice” and impressive image, is at least AI-supported or based on selected AI algorithms.
But I am getting concerned about the option to add details which cannot be there if you calculate the optical performance possibilities of a scope-camera-seeing-darkness combination.
Not that I am against using intelligent “sources” to manipulate images, but as in “normal” photography, I think, it should be mandatory that the creator mentions it in the technical details.
What do you think about this idea? Is it nonsense or a good one?
CS
Georg
hey georg,
I don't have a final personal opinion on the topic, but I'm pretty sure it will turn out to be impossible to force people to be transparent regarding the use of AI tools — or any other tools, gear, or whatever…
makes me think about the vegan police within the scott pilgrim universe :D
This is difficult as with certain tools, it’s hard to know what’s under the hood. All we can do is try to judge by the results. I think you have to draw the line when the results depart from reality. That said, fine structure can radically change in appearance when imaged at low and high resolution. Before you can even make a judgment you need to find a baseline image you can trust (Hubble?). Even then you have to have a good feel for the expected degradation at reduced resolution.
At the end of the day, there is a competitive aspect to the hobby. Even if you don’t expect or care about awards, no one wants to be seen as putting out sub-par work. Even if it’s to just please yourself, every one of us is constantly learning to use the available tools for an improved result. Since we all have access to the same tool sets I don’t think there an explicit need to to say “I used BX or Graxpert”, it’s a given. Now if someone’s ethical boundary includes blending of Hubble details into their image using AI, there’s nothing stopping them from doing it. I would hope that we as group would be able to spot that or at least question the result.
You would hope that anyone doing that would say so but given the motivation, they probably wouldn’t.
Well written Helpful Insightful Respectful Engaging
I used to mention use of all the software and plugins that I used in my software section on my images and many of those were part of a given equipment profile if I used them most of the time.
Just this week I changed that policy so I now just list the basics (NINA, PixInsight, etc.). I did this because there are so many possible processes and probably thousands of related combinations and permutations that just mentioning a software tells the viewer and/or reviewer almost nothing about the real processing and may lead them to think that more of the result was AI than was actually the case. Both the degree to which they are applied and sometimes the order in which they were used vary hugely from one image to the next and often different processes or settings were used on one image component as opposed to another. Saying simply “‘I used it” or “I didn’t use it” can be misleading as a result.
I tend to assume that at least some AI is being used on most images these days unless otherwise stated.
Tony Gondola · Jun 21, 2026, 05:22 PM
Before you can even make a judgment you need to find a baseline image you can trust
Totally agree. In fact, one concern I have about reviews/awards is that I suspect most reviewers do not have the time to do such object based comparisons and might assume that any detail greater than they expected is somehow artificial. As AI advances we will be seeing such “AI Signatures” becoming less and less obvious anyway. In fact, as many are aware, the new Parallax is already less blatant than is BXT.
But even if you do such image comparisons a better image might well be just due to better seeing. I do a lot of planetaries and am often able to produce images that look better than many others not because I am super skilled or because I turned the AI “up to 11” but because I happen to have a remote site that has great seeing a lot. In fact the average seeing on Friday was 1.1 arcsec according to the monitor and the tracking and FWHM showed it. I would hate to think that people believe I am overusing AI when much of the resolution is really the skies and my aggressive rejection of less than ideal subframes.
I would also point out that since good seeing improves contrast and most AI is contrast driven, better seeing can make AI work better and more accurately.
On a peripherally related subject, I noted this from Innovations Foresight this morning. I am not quite clear on the details but it appears to be essentially ”AI AO” with less hardware. If this is correct and this sort of thing does become part of the amateur imaging community it will further “blur” (ironic use of term) the lines as AI will affect the data, not just the final result.
Would love to hear an expert take on this tech (Maybe John Hayes?)
Engaging
Bill McLaughlin · Jun 21, 2026, 06:13 PM
But even if you do such image comparisons a better image might well be just due to better seeing.
A very important point, sometimes it can come down to conditions and technique.