The astrophotography nightscape develops at warp speed: whatever our views of artificial intelligence (AI), it is a powerful tool to enhance image acquisition, processing, and publishing.
Here are some of my experiences:
As noted in ‘about astroanderson.com’, I have been lucky enough to image in both northern and southern hemispheres. During late summer and early autumn of 2026 (southern hemisphere), I learnt that during the winter of 2026 (southern hemisphere), it was likely we would be relocated for my wife’s UN diplomatic post to near the equator. So, I wondered what would be good to capture or re-capture before leaving, knowing that the Carina Nebula Complex, to me the jewel of the southern sky, would, in autumn, be at a time of its highest transit from my location. My imaging equipment included a Sigma 85mm lens, RedCat 71 (focal length 348mm), Celestron EdgeHD 9.25″ (focal length 2350mm), Celestron CGEM mount, 7nm narrowband filters, and ZWO ASI2600MM camera. With my CGEM mount, I have never been very successful with meridian flips and so try to avoid them.
(Being born on 2nd June), I am a fan of Google Gemini – so I gave Gemini the list of my equipment, my location and my goal (to optimize my last chances of imaging Carina Nebula Complex from a prime location), and asked what should I image with my different pieces of equipment and, during April 2026, when could I start imaging as soon as it was sufficiently dark, avoiding meridian flips. After a bit of back-and-forth AI conversation, which I always find productive, Gemini returned its Blueprint of objects, their celestial coordinates, start times and suggested exposure times, which, with three different imaging targets, I could complete over nine nights. I managed this, although had to cut a few corners (reducing imaging time and total exposure time) due to a lack of clear skies.
Here are the results, all of which, of course, could be better; but I was happy to have it done during the last months at my fortunate location.
For the Sigma 85mm lens, Gemini suggested capturing the Running Chicken Nebula (IC 2944), the Statue of Liberty Nebula (RCW57) and the Carina Nebula (NGC 3372) in a row.
I then manipulated a monochrome version with Adobe Photoshop to portray the Carina Nebula chasing after the Running Chicken Nebula.
For the RedCat 71, Gemini suggested capturing the Carina Nebula Complex on its own, as it would very nicely fill the shot.
For the EdgeHD, Gemini suggested trying to go for the Mystic Mountain in with the Carina Nebula, a structure that had been famously captured by the Hubble Space Telescope (HST) in 2010. This turned out to be too ambitious, and I was compromised by overall insufficient exposure time due to lack of clear skies.
Nevertheless, I was able to use the gathered data to present a passable image of the Homunculus Nebula, the hourglass shape surrounding the bright Eta (η) Carinae star system, composed of gas and dust ejected by the Eta (η) Carinae star system during a massive outburst observed between 1838 and 1845, and expanding at 2.1 million km/hour.
Overall, Gemini was creative in offering solutions, responsive to my own suggestions, reassuring in what could potentially be done, and saved me a lot of time and uncertainty in my planning.
Once the all the raw images are integrated with Pixinsight’s weighted or fast batch pre-processing, specialized neural networks and AI tools really came into their own.
Sharpen with BlurXTerminator
This PixInsight plugin utilizes deep-learning models trained on real space telescope data. Instead of relying on traditional deconvolution methods, AI analyzes the image, measures localized distortion, and sharpens details in short periods of time, correcting optical issues like coma, chromatic aberration, and tracking errors across the entire frame. It separates stellar and non-stellar elements to prevent ringing around stars.
Here is an example of how helpful BlurXTerminator can be.
This was the outcome of integrated exposures with the H-alpha filter of the area around the Mystic Mountain in the centre of the Carina Nebula. When I first saw it, I felt I had wasted my time, and that focussing had all gone wrong (Celsteron EdgeHD sometimes being notoriously difficult to focus deep sky objects). But, BlurXTerminator saved the day (or, night of imaging), as can be seen in the resulting image that can be readily further developed – hover over image to see BlurXTerminator result.
One of BlurXTermintor’s strengths is in rescuing potential imaging errors and sub-optimal images to provide something that can be further worked on to produce a satisfactory, if not necessarily ideal, outcome.
Upscale with Topaz Gigapixel AI
When cropping tightly into small targets like the Pillars of Creation in the Eagle Nebula (M16), resolution can be lost and pixelation introduced. I print my images on A3+ Hahnemühle fine art paper mounted on aluminium backed foam boards. To ensure high quality images, I use Topaz’s gigapixel AI deep learning to upscale my images by 200%, with gigapixel intelligently adding realistic detail rather than just multiplying pixels. Breathing new life into albeit quite high-resolution data, sharpening edges and reducing artefacts seamlessly, gigapixel creates crisper prints for larger-format displays.
I have adjusted some of my images with Adobe Photoshop’s Firefly AI engine via Generative Fill and Generative Expand. For example, I captured this image of the Andromeda Galaxy, M31, with narrowband and RGB filters. I like it very much, and nebula can be clearly seen within the galaxy in red. As the Andromeda Galaxy is one of the most frequently imaged objects, it has often been said be creative and present the galaxy in a different way (for example, read Charles Bracken).
When I look at this image of the galaxy, I imagine it on a dramatic journey from bottom left to top right. So, I wanted to adjust the image, placing Andromeda at the bottom left of the image and generative expanding the star background upwards and to the right, also giving me space to rotate the galaxy so it is at 45⁰ across the diagonal of the image canvas. Firefly did this perfectly, intelligently extending the starfield background flawlessly.
For some, adding an additional star background with “Generative Expand” would cross a line from documentary astrophotography into digital art: to me, though, astrophotography is a blend of art, science and technology – within Pixinsight, whenever we use generalized hyperbolic stretch or curves transformation, we are adjusting images to our own taste: I find it acceptable to emphasize artistic intent and creativity over rigid technical replication.
With the help of Gemini, I then boldly went several steps further, perhaps too far, seeing Andromeda on the move, first at impulse power;
and then at warp speed.
As much of the challenge and reward in astrophotography is in acquiring and processing images, for me, it is also in publishing images, particularly on the web. I have hosted and set up my website, astroanderson.com, with Hostinger UK, WordPress and Elementor Pro. Whilst there is a learning curve, much of the heavy lifting for me has been done by Elementor Pro, managing advanced design widgets, theme builders, and form builders. Kodee, Hostinger’s built-in AI assistant is deeply integrated into the Hostinger ecosystem and has helped to build, manage, and troubleshoot astroanderson.com.
However, there are limitations, particularly when trying to ensure seamless working across WordPress, Elementor Pro and a whole range of useful (and, often required) WordPress plug-ins such as SearchWP, which I use to manage searches within astroanderson.com.
At times, Kodee has struggled to deal with this – so, step in Gemini! (I have not tried Elementor AI, as that requires an additional subscription, which, at present, I have decided not to pursue).
When interacting with Gemini, I give it my clear goal, purpose and structure of the website, slugs of website pages (the last part of a URL address that identifies a specific page in human-readable words), and any existing HTML code or custom PHP snippets (including those previously produced by Gemini). Gemini produces high-quality solutions that are tested and adjusted as needed. As the ‘conversation’ is continued, Gemini gets a better understanding of my preferences and goals, producing enhanced code, and suggesting a range of ideas of other things that could be done. I also leaning coding, being then able to make further adjustments without Gemini’s input. When logging off, the conversation is not kept, so Gemini advises what information is needed to continue the conversation another time. I have particularly worked with Gemini in managing searches within astroanderson.com.
The Bottom Line
AI is a tremendous asset across the entire lifecycle of a modern astrophotographer’s workflow. From the initial planning phases in specified locations to the final pixel polish and the web code that publishes images to the world, AI frees up time so less energy is spent struggling with software and more time can be spent contemplating what is out there. The key is balance. Use AI to assist with the heavy lifting and the generation of ideas, but always bring one’s own human creativity, passion, and oversight to the final result.
