Creating an "Arc Shot" from a single photo with AI

I’m a tremendous fan of local AI, especially in creating videos and images. There’s lots of specialist open-source models, and by combining them you can create some really interesting effects.

Let's say we have a photo or an AI-generated image:

Then there's a specialist version of Qwen Image Edit model that can change a camera angle from an image, so we can generate a view of this kitten from 3 other angles:

Then we can then feed these three images into the Wan 2.2 or LTX2 video model as a series of first and last frames for video generation.

You can do all of this on consumer-grade equipment using a piece of software called ComfyUI.

The same process works with an actual photo - so, for example, here's a screenshot I took of Jon Clayton, who recently interviewed me on his Architecture Business Club podcast:

I then ran this through the Qwen Image Edit model to create 3 other angles of the shot and then plugged those into the Wan 2.2 model and created the following video:

The interview I did with Jon was really great too - his show is mainly aimed at Architects but we talk about a lot of other things too here: