The Image Is Easy. The Campaign Is The Work
One convincing frame is a starting point. Making a whole campaign feel like it belongs together takes direction, judgement and a fair bit of fixing.

How it feels, at times -Comfy
Getting one good image out of an AI model can feel like you’ve cracked it.
The lighting works. The composition has something. Nobody has an unexplained extra hand. You save it, send it to someone, and briefly consider yourself unstoppable.
Then you need the next shot.
Same character. Same product. Different angle. And suddenly the jacket has changed, the packaging has invented a new logo, and the room appears to have been renovated between frames.
This is where the production work starts.
Start with something worth making
At Warm Regards, the interesting question isn’t simply whether something can be generated. It’s what we actually want to make, and whether this is the right way to make it.
A strange little film and a commercial campaign might use some of the same tools. They don’t necessarily need the same process.
One might depend on an unexpected moment that makes the whole thing work. The other might require a product to remain recognisable across a dozen deliverables, with room for copy and a deadline that doesn’t care which model launched yesterday.
The brief has to lead. The tools come after.
Give the idea some rules
Before expanding a visual direction, it helps to decide which parts are allowed to change.
Maybe the environments can be wildly different, but the product needs to stay exact. Maybe the camera can move freely, but the lighting should feel like it belongs to the same afternoon.
Those decisions become a practical checklist:
- What must remain consistent?
- What can change between shots?
- Which details will the audience notice first?
- Where does the brand need precision?
- What needs to survive a vertical crop?
These aren’t glamorous questions. They are, however, much easier to answer before producing forty variations of the wrong thing.
A frame can look brilliant on its own and still be wrong for the campaign.
The useful part is often the fixing
There’s a version of AI production that appears to consist entirely of typing a sentence and receiving a finished commercial.
Lovely idea.
In practice, a promising generation might be the beginning of a shot rather than the end of it. An expression works, but the background doesn’t. The movement feels right, but the final second falls apart. A product looks convincing until you compare it with the actual product.
That’s where editing, compositing, retouching and ordinary creative judgement earn their place.
Sometimes the answer is another generation. Sometimes it’s a small manual correction. Sometimes the sensible thing is to build or shoot an element rather than keep asking a model to guess.
There’s no prize for using the fewest tools.
Review the sequence, not just the favourites
A useful review is to put everything next to everything else.
- Look at the images as a set.
- Watch the shots in their intended order.
- Check the actual delivery crops.
- Remove anything that only works in isolation.
That last step can be annoying. Occasionally the best-looking image is the one that doesn’t belong.
Keeping it because it took ages to make doesn’t improve the campaign.
Volume should not mean starting over
Producing more work becomes manageable when the important decisions don’t need to be rediscovered every time.
An agreed visual direction, useful references and a clear approval process give the next deliverable somewhere to start. That leaves more attention for the things that genuinely need a creative decision.
The aim isn’t to make every image identical. It’s to make the differences feel intentional.
A close-up, a wide shot and a vertical cutdown should each do their own job while still feeling like parts of the same idea.
Make something you would actually stop for
Technical consistency matters. So does having a reason for someone to look in the first place.
A production can be beautifully organised and completely forgettable. The process should leave room for humour, tension, an odd detail or an image that makes someone pause.
That’s the balance worth chasing: enough structure to deliver the work, enough freedom to make it interesting.
The tools will change again.
The question stays fairly simple.
Is this worth someone’s attention?
Warm regards
Warm Regards