AI fun - still to video, not yet perfect.

AI can now generate a short video from a single still photograph. LightRoom has inherited Photoshop’s latest feature which creates a three-second clip from an existing image, so I thought I'd see how well it coped with a couple of my River Line photographs, despite all the effort we have put in to rid ourselves of those pesky live videos created by iPhones.

I thought I would give it a go with a photo of a rowing on the river (above).  The result (below) is surprisingly convincing at first glance, but then I realised something was very wrong – they're rowing backwards! The cox is looking where they've been rather than where they're going. Then, looking more closely, the right-hand oars begin to disappear. By the end of the clip only the stroke rower still has one, and that has somehow grown to an enormous size. At least No. 3 has stopped looking over his shoulder and destabilising the boat!

Perhaps rowing crews are simply too difficult for AI. What happens if I use a static subject like a building instead?

The result is rather more successful. I used Lightroom's suggested prompt to create a gentle panning shot. It's an interesting effect, although perhaps a little over-enthusiastic in places. I suspect the key is learning how to write better prompts, but I also have to remember that this is a premium feature and every video uses AI credits, so too much experimentation could become expensive.

It's another reminder of just how quickly AI is developing. The results can be remarkably convincing at first glance, but photographers still have the advantage when it comes to spotting the little details that give the game away.