I had a half-decent play with NanoBanana Pro yesterday. I’m not sure whether it’s Pro 3 or just NanoBanana Pro, but it was released around the same time as Gemini Pro 3.
I have to say that I think we’ve probably arrived at the 95% mark in terms of realism. If you prompt it right, it hits 100%. There’s no way you’d distinguish it from any other photo taken, especially if you mimic slightly older devices in the prompt so that there are imperfections in the photos.
I also had a play with the second generation of Google’s video generator, and while it made some epic mistakes, the basic quality of the video and audio was remarkable. See Google AI for more on Google’s work in this area.
So, in another year, when another 50% of marketing agencies have adopted AI instead of using videographers, editors, set crew, and lighting crew, what happens to those jobs? Are we ready?
Observations:
- Image generation: realism feels like it’s reached the 95% mark—100% with the right prompt and deliberate imperfections.
- Video generation (2nd gen): impressive base quality for both video and audio, but still capable of large, noticeable mistakes.
- Implication: rapid AI adoption in marketing could displace many production roles—videographers, editors, set and lighting crew—within a short timeframe.