My thoughts on Aftershoot for wedding photographers

Is Aftershoot a good tool for wedding photographers?

Artificial intelligence is everywhere now in photography tools: Aftershoot uses AI to cull shoots; Adobe‘s using AI in Lightroom Classic’s masking and content-aware fill tools; Topaz Labs and DxO both use it to power their image quality apps; and ImagenAI uses it to learn your editing style.

Some of these are now a permanent part of my wedding photography workflow – Lightroom’s new masking and healing tools are like magic and save many a diversion over to Photoshop; DxO’s Pure Raw 2 is so good with high ISO files that it’s literally changed my working life for the better; and ImagenAI gets me closer to the finished image than a Lightroom preset so I have more time to focus on creative edits (seriously, I love it – use this link to set up a trial account and they’ll give you an extra 500 credits for free – 1500 free edits in total).

Unfortunately I didn’t really click with Aftershoot – but it’s important to say that is just my opinion, and you may have a completely different experience. Personally, I found learning to use it more frustrating and time consuming than just culling the shoot myself, and I never trusted the results completely anyway.

I don’t think anyone really ‘enjoys’ culling a wedding shoot manually, for all sorts of reasons. For example, many of us photographers suffer from serious ‘imposter syndrome’ and criticise our own work far too harshly, so the culling process opens us up to some serious opportunities to self-criticise!

But in testing this particular AI culling tool I’ve learned I need to know for sure that every image in the gallery is telling the story that I want to tell in a way that best represents both my style and the people in the photos, and leaving it to an AI which I then feel compelled to double-check is a false economy.

Disclaimer: all of this is just my opinion, and lots of photographers claim to love it. So you should probably give it a go just to see how you feel about it.

UPDATE, October 2022: This post was originally written in April 2022. It’s now October 2022 and the hard-working team at Aftershoot have released significant updates based on user feedback. According to this post on their blog several recent updates significantly finesse the interface and how it makes selections, especially with regard to ‘closed eyes’ and what’s considered a duplicate image.

So just to re-iterate my disclaimer in bold above, I really do believe that if you’re at all curious about whether AI culling could work for you, you should give Aftershoot a try. It’s not for me, but you might love it.

So can Aftershoot really cull a wedding for you?

Subjectively speaking, I think you know my answer!

Objectively? Well… I think if I showed you what Aftershoot produced at default settings when I tried it back in April 2022 you’d be impressed at how usable the results seem. On top of that there’s a great range of settings to customise; for example, how many you want to cull, and your threshold for eyes-closed, or out-of-focus images (both of which have apparently had big updates in their effectiveness since I tried it).

Plus the promise of being able to correct it and teach it are exciting. I didn’t think it would be able to take over culling completely, but I hoped it could make some impact.

When I started going through what it rejected, that promise started to lose its shine. I felt I was swapping one manual workflow for another which is arguably less intuitive and more complex than the first, resulting in a final selection I didn’t really ‘believe’ in.

The idea is that you’ll finesse the slider settings and when you make your own final selection it’ll start learning, eventually making ‘better’ choices more often. And I’m sure it will get closer.

But the idea that you could eventually run a wedding through it and barely cast your eye over the results, which is what some photographers are saying they now do, seems like a pipe dream to me. And, to be brutally honest, feels a little lazy. Personally speaking I know I’d always want to check everything, which defeats the purpose somewhat.

The biggest problem I have with AI culling isn’t something that’s easily fixable

Even at the speed with which AI tools are advancing, I don’t believe an AI will ever be able to apply emotional or practical context to the photos it reviews. It can’t possibly know the people or the relationships they have with each other, its choices aren’t story-driven, and they can never be about how a photo makes it feel because outside of science fiction an AI can’t feel.

In fact its choices aren’t really choices, but more the results of a set of very clever filters.

I love the dream it’s pitching, and I wanted it to work out for me but after putting several weddings through Aftershoot I knew I wouldn’t ever trust the choices an AI makes. Ultimately I felt better about culling manually than I did about letting an AI choose for me, especially as I’d feel compelled to check its work every time.

How Aftershoot filters images

The good thing about Aftershoot for wedding photographers is it gives you ample room to tweak its settings and to review everything it’s graded and culled for you. The problem for me was that doing so adds frustration and time to a process that’s supposed to be reducing both.

With ‘sequences’, Aftershoot looks for sets of almost identical images shot close together, bundles them into a group, and picks what it thinks is the best.

One of my issues is that it often creates several different ‘sequences’ when I shot through one moment on the day. Or, it might create one sequence with a single ‘best’ shot, when I may have picked several to tell a story.

Even if the selection process itself is improved (which apparently it has been since I first wrote this) I immediately started falling out of love with the dream. I began to realise at this stage that I just don’t want to learn a whole new system, I just want to whizz through everything I shot in a linear fashion, instead of jumping between sets of ‘sequences’ and trying to compare multiple what the AI thinks is the best from each sequence with all the ‘rejected’ images in another bin.

But the key issue is still the lack of a human element. When I’m selecting the best from a sequence I’m driven by story and emotion – does this image capture the moment better than that one, and does it make me feel something? I found Aftershoot consistently didn’t pick what I’d have picked, and those ‘wrong’ choices weren’t the kind I’d be happy to compromise on. I’m not okay with ‘close enough’.

Image blur & closed eyes

Take my experiences here with a pinch of salt because these features, like duplicates, have apparently been significantly updated. However, I’ve not tested the improvements because I’ve decided that AI culling just isn’t for me. 

The other filters, for blurriness and closed eyes, have great potential to strip out duff shots straight away. But even after I adjusted the settings it wasn’t getting me significantly closer.

The ‘blurriness’ filter flagged tons of photos that simply had very narrow depth of field because I shoot wide open on fast primes for most of the day, and it also found closed eyes where eyes weren’t closed, but rather looking down or away from the camera.

And where eyes were actually closed, it was often people in the background that it was easy to either ignore or crop out, or the eyes were closed for good reason. For example, the bride or groom may close their eyes in portraits together, or during a slow first dance, and Aftershoot flags these images.

Going deeper, it seems to favour people looking at the camera over looking at each other. A lot of people look to the camera when they see it pointed their way, because we’ve spent most of our lives being told to look at the camera and say cheese!

But, I’ll usually prefer shots where they’re looking at each other or their guests, because for me that’s a more emotive image with more human connection, and more in line with my candid approach to wedding photography.

I also don’t want an AI to dump images it deems technically imperfect when there may well be magical moments in there that are worth keeping despite being slightly mis-focussed or motion-blurred. I choose photos like that rarely, but with Aftershoot I’d have to accept most of those photos getting rejected unless I found them myself – which defeats the purpose of the software.

Overall, I found the new workflow (reviewing multiple categories of selections and rejections, then comparing to see if I’d have picked the same) more complex and frustrating than a manual cull making ‘yes’ selections chronologically.

I know the idea is you need to check less thoroughly as it learns what you like to select. And I think for shoots where emotion isn’t as important it’ll really speed up workflow.

But for weddings emotion is important, so I’ll always need to check everything Aftershoot selects closely. So on balance I prefer the simpler, cheaper, and more ‘honest’ method of culling manually.

Am I being too hard on Aftershoot for wedding photographers?

Maybe. It’s made by real people who really believe in their product, so I’ve tried to be clear that these are just my opinions and that you may feel very differently!

And, Aftershoot knows it won’t get it right every time. That’s why they’ve worked hard on tools to finesse its choices. And they’re always making fundamental improvements to the software based on user feedback.

But given that I know I’ll never completely trust any AI to make these choices, why not just cull it myself?

Of course there’s bound to be photographers for whom the limitations I ran into don’t present the same problems. Perhaps the standards I set myself to meet are impossibly high (they probably are). Perhaps studio portrait photographers shooting in controlled environments will love it; their culling likely prioritises sharpness and eye position, and less so the emotion of the image.

A well-known photographer told me they’re “happy enough” with what it selects, that it’s “close enough” for them. The problem is that “close enough” when it comes to the selection of the images themselves isn’t good enough for me for a wedding. That’s no diss to them, everyone shoots differently, sees a wedding differently, focusses on different things in the resulting gallery (no pun intended).

For my style and approach, I prefer to just buckle down and cull the wedding myself. It’s the only way I’ll know for sure that every image in the gallery is telling the story that I want to tell in a way that best represents both my style and the people in the photos.

But if you’re even slightly interested in AI culling, please try Aftershoot for yourself as you may disagree with me completely!