The journey of a photographer discovering AI image generation

Diving into Image Generation

The first thing I tried was pure image generation. My goal was to generate simple landscape photography. If you look at my work, you’ll notice a recurring theme - sunsets.

This preference for sunsets will play an important role in the conclusion. The tool I used for this was the Stable Diffusion variant.

Experimenting with Photo Editing

Next, I decided to experiment with photo editing using the Canva AI. This led to the creation of some impossible, possible, and rare photographs.

The lighthouse image, in particular, would have probably required better tools than I had at my disposal.

However, none of the images created here are beyond the reach of a skilled photo editor.

Trying Out Gen2 from Runway ML

Like many others these days, I decided to try out the awesome Gen2 from Runway ML. I used the text prompt option.

This experimentation led me to create a short film entirely generated with Gen2 from Runway ML.

My take on AI for Photographers

If we consider that photography is the act of capturing the unexpected images the world presents to our eyes, here’s my current take on AI for photographers:

As my sunsets highlighted, to create interesting images, we have to step away from all the images we’ve seen before, otherwise result won't be really interesting
AI often allows only a quicker and lower-quality generation of what a good digital artist, drawer, painter, or photo editor can do.
Creating “new” images that are not redundant and with photorealistic quality requires a lot of prompt engineering and imagination

Given all that, for me, the coming challenge with image generation as an unprofessional photographer will be the following: Emulate the surprise that compels me to take a photograph while maintaining high photorealism. Below are some examples of this surprise: