Miner adjustments

I’ve previously written about how I’m consciously shifting focus in my birdwatching this year, with the goal to deepen my knowledge after a few years of trying to go wide. This extends to my photography and editing practices, and has led to some self-reflection about what I’m trying to achieve with bird photography.

Editing with experience

I’ve spent some time over January reviewing old photos, interested in where I may have pushed the edit too far. I’ve looked at about half a dozen and will likely continue to review more, though only slowly, and only for photos that are worth the time it takes.

This is motivated by my growth as a photographer: I can’t re-take old photos, but with more experience as an editor I can re-examine how I processed them afterwards, and fix what I would now view as mistakes.

That interpretation itself is built on me having a better sense of what I want my photography to convey. When I started shooting birds I used editing to push images as far as possible toward warmth and brightness, even when this was in defiance of the actual conditions. That style is dominant on Instagram in particular, and I adopted it by observation. The only story was ‘look at this bird that I saw’.

More recently I’ve been thinking that I want my photographs to reflect my experience of the bird. To me that means that elements like light and warmth should reflect the moment, rather than imposing golden hour on every scene. It also means being intentional about the priorities in the image: where the bird is, what it’s doing, the story being told, the details that add context and those that just distract.

This approach doesn’t mean leaving images unedited. I’m not fetishing the raw here – I’m very happy to adjust things, I just want to be deliberate about it, and ensure that they’re in the same aesthetic language as the experience itself.

Sacred kingfisher

Below is one example. I took this shot in October 2024 at Dights Falls, and was particularly thrilled with it because it was the first time that I’d had a sacred kingfisher perched at a suitable height on an unusually clean perch. It’s still the only time that has happened for me in the City of Yarra.

This photo was taken at 7.44am, more than an hour after sunrise and so past golden hour. We got no rain that day but it was overcast, and it was cold – the Bureau of Meteorology recorded a low of around 5°C that morning. Both the bird and I were under a light eucalypt canopy – some light was filtering through, but it was certainly a more obscured setting than if we’d been in full sun.

The comparison below has both the unedited photograph and the edit that I made back in 2024.

With all these ‘original’ shots I’ve retained lens corrections and the crop applied to the final image to make comparing them easier, but have otherwise made no adjustments. I’ve noticed that on some of these comparisons the bird appears to subtly shift position in the before and after – I believe that this is due to WordPress compressing and scaling the images. Apologies, I haven’t found a way to prevent this.

Clearly with this edit I introduced a lot of warmth – looking back on it in 2026, far too much. I selected the ‘daylight’ white balance preset, which knocked the temperature up +550 and shifted the tint -4 back toward green – in hindsight not at all a necessary change. I don’t have the original edit history now but I suspect that I used an built-in preset on the rest of the image too, because the curves, mixer and grading are also all adjusted toward suggesting dawn light.

I also applied both AI denoise and remove – more on that later.

This is an important photograph for me and it was top of the list for review. I’ve been re-editing from the original photograph, not trying to adjust the previous output, so I went back to zero with this.

Below I’ve put my recent re-edit (right) against both the original photo and against the 2024 edit.

I’ve increased the temperature (+100), but to a far lesser degree. The exposure is up as well (+0.50), more than on the original edit (+0.40), but by making far fewer interventions in the curves (shadow protection and a slight midtone lift) and color grading (a small bump to highlight saturation), and none with the mixer, the overall impression (to my eyes) is still of a darker, and certainly cooler, image. I’ve retained the AI remove but haven’t used AI denoise.

Side-by-side, I think my intention is clearer. The 2024 edit appears far more interventionist, with a kind of perfect, honeyed warmth that both isn’t reflective of the moment and, I think, makes the whole thing a bit generic. Imposing golden hour where the light in the scene doesn’t otherwise support it always reads as artificial to me. By contrast, the bird in the re-edit looks (to me, at least) more like part of its surroundings.

The original shot was also cropped far more closely to the bird than the re-edit. Close cropping was a consistent theme across many early edits and I’ve largely pulled back. As with the raw, I adopted my final crop in earlier images to make the comparison simpler, but the 2024 crop is below.

The adjusted, zoomed-out crop is far truer to the actual experience of seeing this bird. It was at a good height and on a good perch, but they’re a small bird and it was still 5-6m away. I noticed it because I watched it fly in and land, if it had been sitting still I might not have seen it at all.

Other re-edits

I’ll quickly note a few other edits below that I think also illustrate what I’m trying to achieve.

This bell miner I took at Trin Warren Tam-boore in November 2023. I was only a few months into bird photography then, and I remember being absolutely thrilled about capturing an interesting moment as this honeyeater hangs upside-down and picks at a lerp. There’s some softness on the bird’s face but broadly I think the shot holds up.

For whatever reason I made the opposite choice when editing this photo and shifted it back towards an artificially cool temperature. With my re-edit I’ve tried to return to the original spirit of the photo, lifting the shadows and exposure slightly.

A lot of bird photography sees the subject sitting in an almost field guide pose, in a natural setting, not obscured by any vegetation. That’s not how we generally experience birds though, and it’s not how birds experience the world. I’m highlighting this bell miner shot because I think it’s an interesting pose – upside-down and feeding – that is far more reflective of a moment.

The next example might be a case of me both repairing an over-edit and trying to correct an under-edit. A pair of red-whiskered bulbuls were regularly observed frequenting a site in Fawkner in August 2023. They’re not native to Australia but have a well-established population around Sydney and occasionally turn up in Melbourne. At that time I was only a month into bird photography, a few months into regular birding, and this was the first twitch that I ever made: a dedicated trip to see an otherwise unusual bird for the location. They’re still the only bulbuls I’ve seen in Victoria.

Even more than the kingfisher, my original edit massively increased the warmth. Again, I don’t have the edit trail for the original, but I’m sure that I would have used a Lightroom preset for this and then called it a day.

My 2026 re-edit not only tries to restore the original temperature but is also trying to mitigate that most common problem for new photographers: blown-out highlights. I had no sense that I needed to compensate for the bulbul’s white throat when I first took the image, and the result on the photograph is a glow not only on the neck but down onto the breast and tail as well.

I don’t think this is fully fixable with an edit – you lose too much detail with highlights like this – but I’ve given it a good attempt. The bird is far more legible now, at least, though I’ve sacrificed some of the original image’s contrast to achieve that – the leaves on the hawthorn look comparatively flat.

The bird is in the field guide pose that I described before, but here’s an example where I think the broader scene is key to the story: as the red-whiskered bulbul is an established non-native species in Sydney, the hawthorn is also an introduced species and a declared pest in Victoria. They are common in urban settings and private gardens, and the photo suggests the density at which they’re planted here – there are further plants blurred in both the foreground and the background. Here we have one introduced species sustaining itself on the berries of another, and doubtlessly assisting with seed propagation in the process.

This type of story is another that I want to be able to tell with my bird photography.

AI denoise

In the past I’ve found AI denoising in Lightroom to be irresistable; I’m sure that I’m not alone in this. Applying a small amount before export has become part of my workflow.

Reassessing how I edit has also meant reassessing this tool. In part, this is because I think that AI denoise mitigates risk during the photographic process, especially at capture time. When you’ve got denoise available, you can find yourself in the mindset – consciously or not – of shooting more carelessly or pushing the edit further because any image artefacts can be more easily cleaned up by the software. A less than disciplined approach in the field can be adjusted at home.

I also find myself increasingly uncomfortable with how AI denoise changes the image itself.

See the images above of a female superb fairywren. The left side of the slider is without AI denoise, the right side has it applied with the slider at ’25’. I don’t know what the slider values actually reflect, but I’ve tended to use it in that 15-25 range, so well below the default setting of 50 and maximum at 100.

The most notable effect is the waxy appearance that is created by smoothing out the textures – particularly evident in the loss of feather detail. The top of her head looks texturally soft to me without AI denoise; with it, it looks like it’s made of plastic. I also find that AI denoise creates the impression of a ‘harder’ distinction between subject and background. Of course they’re distinct within the photograph regardless, but the noise softens that barrier in a way that I find pleasing – see the differences in her throat or the top of her head as an example of this. AI imposes a sharper boundary.

There’s also a subtle increase in saturation in the image, especially in the reds – you can see this in her lores and in the browns on the back of her head and shoulder.

I’ve seen examples of birds where the AI denoise, attempting to interpret heavy noise and introduce clarity, will melt feathers together in a way that seems completely artificial. Below I’ve got the original sacred kingfisher edit from 2024, with and without the AI adjustment.

Near the top middle of the image AI denoise has a noticeable impact on making those few golden feathers that overlap onto the blue wing look sharp, but on the wing itself or down on the bird’s belly it has inferred the length or direction of feathers in a way that I just don’t think is realistic – or at least, isn’t supported by the photo that I took.

The kingfisher, too, shifts slightly toward the red channel after denoise: the yellow hues appear to tip slightly away from green toward red, particularly on the breast.

A lot of this is only obvious (if I’m not using that term too generously) with a very close crop. The kingfisher is above and I’ve included the full image of the superb fairywren as well below in order to be honest about this: none of the effects I’m concerned about are particularly visible at the photo’s finished size. My point is about the types of of interference that I’m comfortable making in a photo, and for 2026 at least, I’ve decided that AI denoise falls outside of that.

AI remove

The last element that I’ve been reconsidering is AI remove. This one I have more mixed feelings about: taking some element out is a big intervention into a photograph; at the same time, it can have a very positive effect where it fixes an element that distracts from the subject of the potograph.

This is how I use AI remove, and the kingfisher shot (reproduced below) is a good example of where I think it can work: I removed those background branches because I don’t think they are necessary to the experience of this moment. It might be different if there were more of them or they were thick with leaves and suggested that the bird was not on an open perch but was in a densely vegetated area, but that’s not what was happening. I view them as incidental in the scene and, therefore, as justifiably removeable.

I’ll note explicitly too that I haven’t attempted to remove the branch that is partially covering the bird’s wing. This is a red line for me: I don’t remove anything that impacts the bird itself. I’ve no doubt that Lightroom could generate an approximation of the wing that is behind that stick based on whatever training data it has, but that then shifts into a level of dishonesty that I’m not comfortable with.

Here’s another example of how I use AI remove, on a photo from this past week. This little corella was out feeding on Corben Ovals, my first opportunity to shoot a little corella within the City of Yarra. In the original image the bird was feeding on the yellowed grass, with two dead leaves directly in front of and behind it. I felt that these very slightly detracted from the image, so I removed them.

These two examples reflect how I’m thinking about using AI tools in my photography, where it’s not the scale of the intervention so much as what it changes. My bird photography is aesthetic rather than documentary, and I’m trying to emphasise an experiential or emotional truth about a moment. With this approach I see a difference between changes to the subject of the photo as compared to the rest of the image. A tiny shift in the hue of a superb fairywren’s lores, almost imperceptible at full size, is too much for me, removing a whole branch from the background is fine because it clarifies rather than obfuscates the moment.

This principle means that my attitude toward acceptable AI remove is contextually dependent on the photograph. Vegetation isn’t inherently expendable: it would be crucial in a habitat shot, and in the bulbul photograph the hawthorn bush is key to the story.

Conclusion

I don’t expect this approach to be permanent, nor do I see it as a template for anyone else to follow. Any review of my recent photography will also reveal that I still can’t resist the field guide pose on an open perch when it presents itself.

There are also clear situations where tools like AI denoise are genuinely powerful. Low light, long focal lengths and small, fast birds are a difficult combination, and denoise can rescue images that would otherwise be lost. It can produce backgrounds and separations that are visually beautiful, and I’ve used it enough to understand why it’s become so common.

For the moment, though, I’m more interested in edits that preserve how an encounter felt than in maximising what an image can be made to look like. That means accepting noise, awkward light, partial obstruction and distance when those things were part of the experience of seeing the bird. It also means being selective about where I intervene: clarifying a scene by removing incidental distractions feels consistent with that goal; altering the bird itself does not.

I’m not treating the raw file as a kind of truth against which all edits should be judged. An unedited photograph is already an interpretation, shaped by the camera, the lens, and my choices at the time of capture. I’m just trying to be as deliberate in the choices that I make afterwards.