AI Video Generation in Commercial Production: What Works, What Doesn’t
There’s a lot of noise right now about AI video generation “replacing” commercial production. Most of it comes from people who’ve generated a few impressive clips for social media and never had to deliver a paid campaign against a client’s brand guidelines, deadline, and revision list.
We’re not in that position. We use Seedance, Kling, and Nano Banana on real client work every week, alongside traditional CGI. So instead of a hype piece or a doom piece, here’s the actual state of things — where these tools deliver on commercial work, and where they still don’t.
What works, genuinely
Short, contained camera moves on a locked asset. A push-in on a product, a slow orbit, a light sweep, a subtle parallax between foreground and background. When the shot is short — a few seconds — and the subject doesn’t need to change shape or interact with anything else, current tools handle this well. Well enough that we ship it.
Mood and atmosphere shots. Steam rising off food, condensation forming on a cold drink, ambient background motion in a lifestyle scene — anything where the exact physics don’t need to be perfect, just believable. AI is genuinely strong here, often faster than building an equivalent simulation from scratch.
Rapid concept testing. Before committing production budget to a direction, we can generate rough motion tests to show a client “here’s roughly how this would move” days before a single frame of final CGI gets built. This has changed how early client conversations happen — more visual, less descriptive.
What doesn’t work, honestly
Consistency across a longer sequence. Ask an AI video model to hold a product’s exact shape, label, and proportions consistently across 10+ seconds of footage with camera movement, and it starts to drift. A logo shifts slightly. A cap shape rounds off. For anything under a couple of seconds this is barely visible. For anything longer, on a real brand asset, it’s a problem you can’t ship.
Complex physical interactions. Two products touching, a hand interacting with packaging, liquid pouring into a container and splashing correctly — these involve physics that current AI video models still fake unconvincingly under close inspection. This remains simulation and traditional 3D territory, and will for a while yet.
Exact color and material accuracy over time. A brand’s red needs to be the same red in frame one and frame ninety. AI video generation, especially across longer clips, can shift color subtly frame to frame in ways a human eye picks up even if they can’t say exactly what changed. For hero brand assets, that’s not an acceptable margin.
Text and fine label detail. Anything requiring legible text on packaging — ingredient callouts, logos with fine detail — still tends to warp or blur under AI video generation. If a shot needs the label to actually be readable, this is a hard no right now.
Why the honest answer matters more than the hype
We’ve had clients come to us assuming AI video means faster and cheaper across the board. Sometimes it does. For pre-viz, mood shots, and short controlled motion, it genuinely speeds things up and cuts cost. For anything requiring brand-exact consistency over a full commercial’s runtime, it doesn’t — not yet, not in a way we’d put our name on for a paid campaign.
The studios overselling AI video right now are setting clients up for a bad surprise mid-project, when the “quick AI version” doesn’t hold up under actual brand scrutiny and everyone’s back to square one, except now behind schedule.
Where this is heading
These tools are improving fast — faster than almost anything else in this industry right now. The list of “what doesn’t work” above is shorter than it was a year ago, and it’ll keep shrinking. But right now, today, the honest position for anyone doing commercial work is: use AI video where it’s actually reliable, and don’t pretend it’s ready for the parts it isn’t. Clients don’t remember which parts of a project used which tool. They remember whether the final film looked right.
Squirrel Studio combines AI video tools and traditional CGI production for FMCG and F&B brands across Egypt and the GCC, depending on what each shot actually needs. If you want a straight answer on what’s realistic for your next project, get in touch.