Image Generation in Practice: Observations from Real-World Production

Original 3D archviz below, and AI optimized version up top.

AI-based image generation has in recent months gone from being experimental to becoming a real production tool for us at Maestro Media. While a lot is now possible, it should still not be considered a replacement for photography and 3D, but rather a layer on top of existing deliverables. We can reduce time spent on core 3D production and, in addition, achieve very strong results by continuing the process with AI from that point.

In this article, I share concrete observations related to:

• changing seasons in photography

• generating variants of interior scenes

• adjusting camera angles and framing

• implementing specific mood and atmosphere in existing 3D illustrations

• and not least, rights and licensing

All based on real-world use in content production for real estate marketing.

Shit in – shit out still applies in AI image generation

This is perhaps the most important insight.

In our experience, AI is far better at amplifying strong starting points than at improving weak ones. The quality of:

• photography

• lighting

• composition

• 3D geometry

• material definition

determines the final result.

Poor input produces poor output. Always.

Strong original production, on the other hand, enables a wide range of usable variations. This is where the real value lies.

Seasonal changes and interior variants: what works best?

The first image shows drone photography delivered by the client, intended for integrating 3D renderings (the plot), ready for post-production.

The next image shows processing through a standardized prompt we have developed for this purpose.

The final image shows an AI-generated alternative angle that provides the basis for a sales-oriented 3D scene, without the need to communicate the surrounding location.

In practice, we see that AI performs very well for:

• changing seasons in existing photos (winter to summer, autumn to spring)

• variations in furniture, colors, and materiality in interiors

• adjusting lighting, time of day, and atmosphere

• adapting visual expression to different target audiences

• adjusted camera angles

What does not work well is attempting to “rescue” images that were never good enough to begin with. Sometimes it works, other times you end up in an AI-generated whirlpool. A lottery, in other words. At least for now.

In this context, AI is a tool for further development.

GPT-5.2 and Gemini 3 / NanoBanana Pro complement each other

We use several tools in parallel. What started as experimentation is now fully intentional.

• GPT is strong in structure, precise instructions, and control

• NanoBanana often delivers more visual creativity and unexpected, yet usable results

• The same prompt frequently produces different expressions

Results vary, but we clearly experience that combining tools provides far better control than relying on a single one.

This is not about which tool is “best”, but about how they are used together.

Firefly delivers the finish we are looking for

When we add Firefly in the final stage, a clear pattern emerges:

• more consistent quality

• more predictable visual output

• crisper results and higher resolution

• better commercial finish

Time and again, Firefly is what ensures that the result is truly market-ready, not just visually interesting.

Rights and licensing are critically important

Maestro Media AS – AI generated

We believe this is an area the industry underestimates. AI generation builds on existing productions. If you do not have:

• ownership of delivered assets

• the right to further modification

• the right to generate alternative expressions

you effectively lock yourself out of future value creation.

The consequences are serious:

• you may lose the right to further develop your own images

• you may become dependent on new production every time

• you reduce both the lifespan and value of your material

Without clear agreements and license limitations, we effectively pull the rug from under ourselves.

Summary

Maestro Media AS – AI generated

In our work, we see that AI presupposes strong visual craftsmanship rather than replacing it.

Those who succeed going forward are the ones who:

• produce solid original photography and 3D

• understand the limitations of AI

• combine multiple tools deliberately

• and take ownership of rights and future use

Everyone else will produce more variations faster, but without increased value.

Frequently Asked Questions

Can AI change seasons in existing images?

Yes, very effectively, provided the original image quality is high and prompt discipline is sufficient.

Can AI create multiple interior variants of the same room?

Yes. This is one of the highest-value applications in real estate projects.

Is AI sufficient on its own without photography and 3D?

No. AI works best as an extension of professional source material.

Who owns AI-generated variants of existing images?

This depends on licensing and usage agreements and must be clarified explicitly.

Written by Stian Andreassen, SEO/GEO opt. and translation by AI