Today’s newsletter is a little different.

For the first time since I started writing these letters, I’m handing it over to a guest writer. And fair warning: this is probably one of the more technical deep dives you’ll receive from me. It is also a long one, so if you’re interested in getting properly into it, I recommend saving it for a moment when you can sit down and take your time. Mikel has done an incredible job of taking something extraordinarily complex and making it as digestible—and entertaining—as possible.

But before I hand things over to him, I want to explain why.

I first met Mikel Iraeta / BegiBakar in Peru in 2023, and I remember one of our early conversations surprisingly well. At some point, we were talking about photographs we had captured and he casually told me that he could give me his RAW files.

I was genuinely surprised.

Until then, I had mostly encountered photographers who treated their RAW files almost as sacred objects; something intensely personal that you simply didn’t give to anyone else. I had always felt that I would happily share mine with someone I trusted, particularly if we had experienced and photographed something together, but Mikel was the first photographer I met who seemed to look at it in exactly the same way.

And I’ve always liked that philosophy.

If two photographers stand in the same place, that doesn’t mean they’ll come home with the same photograph. We notice different things, make different decisions and ultimately have completely different ways of interpreting the material. If there is trust, transparency about who created what, and an understanding of what you’re both trying to achieve, I don’t think a RAW file necessarily needs to be this untouchable thing.

Sometimes it can actually become the beginning of something neither person would have created alone.

This image is probably the best proof of that.

I captured the RAW material during totality on August 12. But what you see above is the result of handing those files to Mikel and essentially saying: let’s see what you can do with them.

And he took them somewhere I simply couldn’t have imagined possible.

The final artwork is as much his as it is mine: my capture and preparation combined with his extraordinary post-processing work. It is the first time I’ve approached an image this way, and I love that the result genuinely belongs to both of us rather than either one simply assisting the other.

Mikel is also one of those photographers who I think deserves far more people looking at his work. His photography moves remarkably easily between astrophotography and landscapes, cityscapes, food, portraits and completely different genres, and underneath all of it is the same meticulousness and curiosity that you’ll very quickly recognise in what follows.

So rather than having me attempt to explain what he did to my files, it made much more sense to give him the newsletter.

What follows is Mikel’s complete account of how 108 eclipse frames, 12 exposure brackets, calibration data and a rather frightening amount of processing eventually became the photograph above. It goes from the RAW files through calibration, alignment and stacking, all the way to extracting the corona, earthshine, prominences, Baily’s beads and stars before assembling the final image. 

It’s technical. It’s detailed. It’s occasionally wonderfully nerdy.

And if you’ve ever wondered what actually happens between capturing an eclipse and producing one of those extraordinarily detailed images of the solar corona, I think you’re going to enjoy this one.

Over to you, Mikel.

Starting a text with a quote has always struck me as pretentious, furthermore, as something done by those who pretend to appear smarter than they actually are. So how am I going to start? With a quote 🙂

It is a quote that the character Sean Maguire (played by Robin Williams) recites in the film (which I highly recommend) 'Good Will Hunting':

It is a phrase I have thought about a lot since August 12th, and one that I think describes very well how I felt. Months of preparation in which one reads, learns, memorizes, and studies everything a non-teenage human brain is capable of absorbing about the solar eclipse. And yet, none of it accurately describes what is perceived when you witness what is probably the greatest cosmic coincidence in our solar system. No matter how much you read, nothing describes how that made you feel.

So, indeed, I am here to talk to you about the solar eclipse. Specifically, about the workflow that should take you from the RAW files captured during totality all the way to the final image. This is the continuation of the newsletter that Angel Fux sent you a few weeks ago, in which she covered everything related to the capture. Angel very kindly opened the doors of her house to allow me to speak to you about post-processing, so I hope to live up to the occasion.

Let's dive in.

There are many ways to photograph an eclipse, all equally valid. My recommendation is that you use whichever method allows you to enjoy such a singular event. However, if you feel like a bit of rock 'n' roll and are curious to know how to achieve an image that shows the solar corona in intricate detail, fasten your seatbelt and join me down this rabbit hole.

I don't think I am lying if I state that a few months ago I was completely ignorant in this matter, and I would have been incapable of reaching the result I will show you at the end. The experience I gained during the process does not make me an expert, in fact I am far from being one, but it might have a silver lining (which is probably the only reason I would recommend reading this text): I am able to put myself in the shoes of someone facing this for the first time. I understand how intimidating the process is and how overwhelming the sheer volume of questions that arise along the way can be. Do not despair; I encourage you to try. My intention is to write the text that I would have liked to read the first time I faced this challenge. Therefore, I repeat: you will not find an expert here, but rather someone who, after a few bumps along the road, attempts to share their route to show you where to go and why, but above all, where NOT to go.

To facilitate reading, I am structuring this letter as follows:

Two pieces of advice before starting:

A solar eclipse is probably the most complicated astronomical event to photograph and process for several reasons. Regarding post-processing, the most prominent challenges are:

The workflow we are going to follow comprises the following main stages: Calibration, Alignment, Stacking, Processing, and Final Composite.

The first thing I want to make clear is that this step is not strictly mandatory. A solar eclipse image can be processed without using calibration frames. The reason we perform this step is simply that we obtain a superior result when we use them.

To explain why, I will make a small digression to introduce a concept that will be very useful: we are going to work in what is known as linear format. Some of you may know what this means, while others may not. I will summarize it so we all start from the same baseline:

Now that we all know what linear format is, let's explain why it matters. A solar eclipse is a phenomenon with an extreme dynamic range in which the region immediately adjacent to the lunar limb is thousands of times brighter than the outer corona. Working in linear space preserves intact data, allowing us to decide a posteriori how to treat the image. This is much harder once linearity is broken, making artifacts, noise, and banding far more likely to appear during image manipulation.

However, because linear format faithfully and neutrally transmits all information captured by our sensor, it also passes along all the physical imperfections in our optical measurement chain. Therefore, calibration frames are used to correct these imperfections:

In our case, flat frames are common to all shots because the aperture was kept constant throughout the entire acquisition process (f/11). This is not the case for darks, for which we have a dedicated series corresponding to each exposure time. Calibration of LIGHT frames must be performed using the flats, bias, and darks matching their specific exposure time. Several software options exist for calibration; in my case, the tool used was PixInsight, which outputs 32-bit files in TIFF format.

With the files properly calibrated, we can proceed to the next stage.

When taking landscape photographs where elements are static, combining different exposures into a single image is relatively simple because there is no relative motion between components (or it is negligible). This is not the case during a solar eclipse, where relative motion exists between the Moon and the Sun. This poses a fundamental challenge: we must define the reference against which we want to align our images:

This represents an unsolvable conflict, where a choice needs to be made. In our case, the selected reference is the solar corona, since exposure bracketing was performed specifically to highlight its structure. That structure is finely dictated by the Sun's magnetic field and consists of highly characteristic features such as coronal loops (bright arches clinging to the solar limb), polar plumes, and streamers—filamentous structures forming its outermost reach. The key to excellent solar alignment is using these structures as pattern templates so that loops and filaments line up to the exact millimeter.

In this step, we temporarily disregard the Moon, which will be processed separately and composited in a later phase.

There is an additional hurdle: the exposure differences between various shots of the solar corona. A short exposure cleanly highlights structures located near the lunar limb, while in a long exposure that same region is completely blown out (saturated) while revealing outer structures distant from the solar surface that are invisible in shorter frames. Image processing algorithms struggle to find common patterns across images that appear to depict entirely different objects.

As with almost everything, there is no single way to do things, so I will mention the two methods I know can lead you to the final result:

In our case, we used 12 exposure brackets consisting of 9 shots each with a 1 EV step, totaling 108 frames to align. Performing the alignment manually is possible, but I can assure you that by the third bracket, you will prefer rubbing your corneas with pumice stone over continuing. It is an excruciatingly tedious process.

Therefore, Option B was selected. If you didn't understand the technical mathematical explanation above, don't worry, it isn't necessary. It is enough to understand that algorithms (essentially coded instructions) involving a heavy amount of mathematics allow us to align our images at a sub-pixel level. This is a complex enough subject that some researchers have dedicated their PHDs studying and advancing these techniques (see References section). The good news is that we are in 2026, and far smarter people than me have generously compiled all these algorithms into tools available for the rest of us. In this case, you simply access the GitHub repository linked at the end and use the free Anaconda distribution to run tse-tools, a Python-based utility suite. This is the exact tool I used, and the result is remarkably good. I am aware using command-line tools can feel intimidating, but you can configure it in a couple of hours, saving you dozens of labor hours. My eternal gratitude to the developer who made this tool available for the common good.

Before wrapping up this section, a question might be on your mind: How do I know if my alignment is accurate? Photoshop offers a very practical feature to verify this. If you select the upper layer you wish to align against the base layer and temporarily set its blend mode to Difference, perfectly matching regions will turn pitch black. This allows for a swift and visual assessment of alignment accuracy.

With the images aligned, the next step is stacking, where exposure sequences are combined into a single final image with a vastly expanded dynamic range. We execute this in two stages: first, we merge the 9 exposures of each bracket to create an HDR file consolidating exposures ranging from 1/125s to 2s. Since 12 brackets were shot during totality, we generate 12 such HDR files. Next, we stack those 12 high-dynamic-range images into one final consolidated dataset. We do this to boost the signal-to-noise ratio (SNR) of the final image, producing cleaner files with significantly reduced noise. Below is a diagram illustrating this workflow:

As in the previous section, there are multiple ways to execute this, and I understand many might be tempted to use Adobe Camera RAW's 'Merge to HDR' function. Do NOT do this, it will destroy your image.

Before explaining the stacking strategies, it helps to illustrate why combining different eclipse exposures is so tricky. Let's look at the following image:

As mentioned, the corona exhibits extreme luminosity drop-off between its outermost regions and those adjacent to the lunar limb. Each exposure in a bracket properly exposes a specific zone of the corona, leaving under-exposed (dark, noisy) and over-exposed (clipped) regions on either side. The cleanly exposed band varies per frame. Yet the corona is a continuous entity with no hard breaks; thus, blending these exposures naturally using only valid data from each is far from trivial. Indeed, it would be impossible without prior razor-sharp alignment.

The optimal way to combine frames is through mathematics, merging all exposures in linear format. Because linear frames maintain direct proportionality between sensor excitation and image brightness, blending is mathematically simplified to counting photons in linear space. The tse-tools software allows completing this step to produce a 32-bit linear output. However, I won't delve too deeply here because it led me down a dead end, not because the tool fails (on the contrary, it works wonderfully), but because I was unable to achieve a visually satisfying result. Human vision perceives light logarithmically rather than linearly, requiring a translation of the linear file into human perceptual space (a process known as 'stretching'). I failed to produce a satisfactory stretch through this pipeline, which is entirely my own limitation.

So I will focus on the method that DID work for me, for which I must thank master astrophotographer Miguel Claro, whom I had the privilege to meet during the eclipse. The primary objective of this method is preserving natural visual aesthetics when processing solar eclipse images. The procedure begins by selecting an untouched intermediate reference frame (in this case, the 1s frame) exported directly from Camera RAW (crucial: do not touch any sliders). This exported ACR file serves as a visual benchmark in PixInsight to define the non-linear stretch curve applied to the previously calibrated and aligned linear files. The aim of the applied stretch is to match the visual appearance of the linear file to the ACR export—a key step to ensure stretching looks natural to the human eye. Once this transformation curve is established, identical mathematical transformations are applied across all remaining bracket frames. This uniform extrapolation is fundamental: it preserves the physical hierarchy of relative brightness across exposures (preventing non coherent stretching between individual exposure frames) and ensures a seamless light intensity progression across the entire sequence.

Once stretched, the images are exported as 16-bit TIFFs to Photoshop, where they are stacked inside a Smart Object using ‘Mean’ stack mode. This is a mathematically sub-optimal compromise regarding pure local image quality, as pristine pixels are 'polluted' by pixels from frames that are blown out or black in those same regions. However, the perceived result is highly satisfying to the human eye. Mean stacking smoothly compresses the dynamic range, yielding a seamless radial gradient across the solar corona without hard edges or artifacts.

This process is repeated for each of the 12 exposure brackets, which are then combined into a single final stacked dataset. Signal accumulation produces a high signal-to-noise ratio despite the sub-optimal stacking method. The entire process yields an image like this:

Now at last, we can proceed to Photoshop.

If you've made it this far, thank you for persevering. At this point, I had been working on the image for nearly a week, generated almost 1TB of temporary files, and hadn't even opened Photoshop yet! You know that feeling when you sense things are getting slightly out of hand? That was on Tuesday, and we were already at Saturday. Courage, we're almost there.

In Photoshop, we divide the work into distinct stages, listed in order of priority:

Upon importing the stacked HDR result, we typically end up with a low-contrast image where fine coronal detail is invisible. Solar corona processing focuses on enhancing structural contrast and pulling out as much fine detail as possible.

A well-known technique for local contrast enhancement is Photoshop's Unsharp Mask. These tools isolate high-frequency details by subtracting a low-pass filtered (blurred) version of the image from the original. The high-pass detail layer can then be blended back using various blend modes.

However, these standard tools are optimized for Cartesian coordinates (X, Y), which is ill-suited for processing a solar corona. Standard tools fail because they apply a 2D Gaussian blur that averages pixels equally in all directions, blending blindingly bright inner light with dark outer penumbra. By failing to respect the radial geometry of the Sun, the algorithm mistakes natural gradient drop-off for a sharp contrast edge. Subtracting this blurred image generates severe dark halos around the lunar disk and dark artifacts where light intensity drops off sharply. To remedy this, it is recommended to work in polar coordinates* (r, φ), which naturally conform to solar corona geometry by respecting radial filament structures and assuming roughly equal brightness for pixels equidistant from the solar center.

The question now is how to perform a coordinate transformation in Photoshop. One option is writing custom Python code, but at this stage of the process, drinking hemlock sounds more enticing. Fortunately, Photoshop provides two built-in filters that produce results equivalent to applying unsharp masking in polar coordinates: Radial Blur & Zoom Blur filters.

Before applying either tool, ensure the Sun is centered perfectly in the frame. If not, align it to the exact center first.

Once inner and outer coronal details are extracted, our task is blending them seamlessly to create a visually appealing, cohesive transition. Throughout this phase, we have ignored the Moon, which will be processed separately and blended later.

The Moon requires its own workflow, similar to (though simpler than) that of the corona. The goal here is revealing a crisp, well-defined lunar disk with enough extracted shadow detail to showcase earthshine.

Recall that initial alignment was referenced to the solar corona, which causes lunar edges to blur significantly due to relative motion between the Moon and Sun. To obtain a sharp lunar perimeter, alignment must be repeated referencing the Moon. This will blur coronal structures, but that is irrelevant since the Moon is our target here. Unlike corona processing, we only stack 1s and 2s exposure frames, as these contain sufficient signal to reveal the earthshine.

Alignment can again be performed manually or via tse-tools (the latter is recommended). Stretching follows the same procedure as for the corona, followed by stacking to produce a single image from all 1s and 2s frames (36 seconds total integration time).

Extracting surface detail from the dark Moon is tricky, and I again thank Miguel Claro for his guidance during the Capture the Atlas event. First, we import the file into PixInsight and apply the LocalHistogramEqualization process, which magically unveils underlying lunar topography. Next, we bring the file into Photoshop to enhance structure and contrast.

Once the Moon looks right, we blend the Moon and the solar corona to achieve a balanced result.

The image contains almost everything we want, and we already could stop here. However, it would be a shame not to leverage the treasure trove of RAW files (thanks to Angel) to add a detail that enriches the final result: solar prominences visible during totality, adding a welcome burst of vibrant color.

To do this, we select frames where prominences are sharply defined (or a combination to align and reduce noise). We process the RAW frame to enhance brightness and saturation so prominences pop. The result is imported as a layer in Photoshop, using a luminosity mask to refine the selection and blend only the desired features.

In this case, I added prominences on both sides: a prominent eruptive feature on the left visible to the naked eye during totality (a stunning sight!), and on the right, alongside another prominence, I blended a frame capturing Baily's beads (a nice touch to set this image apart).

Finally, we add stars. Although the eclipse occurs during daytime, ambient darkness drops dramatically during totality, allowing background stars to be captured.

Stars can be integrated from longer exposure frames using blend modes. In my case, corona-aligned HDR composites already revealed star positions. However, because they were aligned to the corona, stars appeared as double or triple dots. Knowing their true positions, I cloned out the trailed stars and painted them back on a new layer point-by-point using custom brushes with varying sizes, hardness, and opacity. While unconventional, this offered complete control to achieve a convincing look.

With all elements composite-blended, final adjustments for contrast, brightness, and color balance are applied.

In this case, since the eclipse occurred at an elevation of only 8° above the horizon during totality, its corona exhibits a significantly warmer hue compared to high-altitude midday eclipses. This signature warm tone makes the image stand out from typical eclipse photos online; however, I wanted to keep the warmth under control. Conversely, I applied a subtle desaturated blue tint to the Moon to contrast against the warm corona, enhancing visual appeal. Lastly, the subtle color accents from prominences break up the chromatic monotony, giving the eye a focal highlight.

Here is the final result:

I hope you enjoy it.

As much as the internet can sometimes seem like a viper's nest, you also meet genuinely good people. The result you see wouldn't have been possible without help from many individuals whom I'd like to thank for making my life easier. Searching for eclipse photos, I came across @fer_astrolandscape_photo, who besides posting awesome photos took the time to point me toward Iker Zubizarreta. Iker kindly answered my questions and shared the repository containing the alignment software I used, along with a user guide he authored. Furthermore, though he doesn't know me, credit goes to Samuli Vuorinen, who compiled Dr. Druckmüller's thesis code into a Python application accessible to us mortals. (Python toolsuite for total solar eclipse image alignment and processing: https://github.com/naavis/eclipsetools/)

I would also like to thank the entire Capture the Atlas team for organizing such a seamless event where everything went off without a hitch. I can't imagine the countless hours behind organizing an event of this scale, but their hard work allowed attendees to focus on what they love most: taking photos and sharing them with friends.

I must not forget Miguel Claro, a guest expert on the Capture the Atlas panel and my ultimate benchmark for eclipse astrophotography. Following Miguel's guidance, I learned techniques I didn't even know were possible. If you aren't familiar with his work, check it out, it is well worth it.

Lastly, I want to express my deepest gratitude to Angel Fux, without whom this photo would simply not exist. The RAW files she captured were flawless. While it is easy to judge efforts solely by final results, the massive time investment makes my gratitude non-negotiable. Success lies in the details and while the rest of us were celebrating totality, Angel extended her session to capture those calibration frames that most people forget. Her discipline and attention to detail made this photo possible—a photo that is as much hers as it is mine. Thank you, Angel.

Druckmüller, M., Rušin, V., & Minarovjech, M. (2006) “A New Numerical Method of Total Solar Eclipse Photography Processing,” Contributions of the Astronomical Observatory Skalnaté Pleso, 36, 131–148.

Druckmüller, M. (2009) “Phase Correlation Method for the Alignment of Total Solar Eclipse Images,” The Astrophysical Journal, 706, 1605–1608. doi:10.1088/0004-637X/706/2/1605

Druckmüllerová, H., Morgan, H., & Habbal, S. R. (2011) “Enhancing Coronal Structures with the Fourier Normalizing-Radial-Graded Filter,” The Astrophysical Journal, 737, 88. doi:10.1088/0004-637X/737/2/88

Druckmüllerová, H. (2014) “Application of Adaptive Filters in Processing of Solar Corona Images,” PhD thesis, Brno University of Technology, Faculty of Mechanical Engineering.

If you’ve made it all the way down here, first of all, thank you. I know this was a particularly technical one, and probably quite different from the letters I usually send. It certainly won’t be for everyone, but I hope that for those of you who wanted to go down the rabbit hole with Mikel, there was something useful to take away from it.

And once again, a huge thank you to Mikel for taking the time to put all of this together and for sharing the process so openly.

I’d also genuinely love to hear what you thought of this slightly different format. If you enjoyed having a guest writer here—or would like to see more letters like this in the future—feel free to send me your feedback.

Until next time, clear skies.

Angel