I tried to get Veo to replace our designer. Here's what actually happened.

Firstly, yes, I am the designer in question!

I started my career as a graphic designer. Then years as a 2D animator: mascots waving, icons lighting up, a brand's whole personality distilled into eight seconds of motion. Clean, friendly corporate work. You've seen a thousand versions of it without noticing.

I love that work. I also love tech. Right now those two things are in tension, because the tech is starting to come for the job.

So when Google AI Studio launched Veo 3.1, which started producing genuinely convincing video, I didn't want someone else's verdict on it. I wanted to test it against the exact work I used to do by hand and see how close it actually gets. Not a hot take. A real test, from someone who knows both sides well enough to judge it fairly.

Two separate questions. Could Veo generate our own on-brand creative well enough to replace a designer in the workflow. Could it produce photorealistic, documentary-style footage for ad concepts, the kind we'd otherwise buy as stock video.

Round 1: an open brief, no guardrails

‍

‍

I gave it almost nothing, deliberately, to see what it would invent:

"Snifferdog is an AI-powered advertising platform that automates campaign launch and management across Google, Meta and TikTok for small businesses, replacing hours of manual setup with a few minutes of automated work. Generate an 8-second advert for Snifferdog."

What came back was a magical beagle with a garbled bandana, using a computer in an office that looked like it belonged in a liminal-space horror short. Unusable. Also a useful reminder: "AI-powered advertising platform" means nothing visually unless you tell it what it should mean.

Round 2: giving it Tove

‍

‍

Tove is our mascot, a robot dog I designed and have personally animated more times than our developers have been able to build features for, if you ask them. I gave Veo a still of her as the starting frame, a proper brief this time, and a tagline:

"Make this image of our mascot, Tove, into a short animation advertising Snifferdog... Keep the animation friendly and futuristic. The tagline is 'setting up ad campaigns should be easy.'"

Tove sprouted legs. Appeared to become possessed. Started chanting in an American accent against a backdrop of muddled logos. Somehow worse than the beagle.

Round 3: getting specific

‍

‍

This is where it started to work. The difference came entirely from how much room I removed for the model to improvise. Instead of a creative brief, I wrote something closer to an animation exposure sheet, second by second:

"Using the provided image as the starting frame: Tove, a white robot dog mascot with a navy visor face and teal eyes, stands centre frame against a deep navy studio background... Three small rounded icon tiles float in front of her at chest height, in a row... She raises one paw and taps the pink tile; it lights up bright and pulses once with a soft chime. She taps the second tile; it lights up. She taps the third; it lights up. By the sixth second all tiles glow steadily in sync... Tove tilts her head slightly and gives a small satisfied nod. Static camera, no camera movement, clean flat character-animation style matching the mascot's original illustration style..."

The result is genuinely cute. Not perfect. Tove's design drifts slightly through the clip and she picks up eyebrows and slightly anime-esque eyes that aren't in the original character sheet. It doesn't give me nightmares. With some tweaking, it's usable.

The pattern held throughout: the less room I gave Veo to interpret, the better it performed. That's not a creative tool behaving like a creative collaborator. It's a literal-minded animator that does exactly what you say and nothing you didn't.

Text is where it falls apart

Anything beyond one or two words came out jumbled, warped, or faintly unsettling. Consistently, across every test. If your creative depends on legible on-screen text, Veo isn't there yet. Full stop.

Round 4: the photorealistic side

‍

‍

Our "there are better things your team could be doing right now" concept was originally built around busy animals: a bee, a beaver, a dung beetle, leafcutter ants. Productive creatures doing productive things. The obvious trap: if I just want footage of a beaver being a beaver, I can buy that from a stock library for a fraction of the effort. So the test was whether Veo could do something a stock clip couldn't. A David Attenborough-style continuous sequence, moving between animals in a single take.

The result was nice enough. It also screamed "AI" in a way the Tove animation didn't. The transition between animals was technically impressive. I didn't love the overall feel.

Round 5: It's like herding cats

Next, for "your team's Friday shouldn't feel like this": cowboys in the Wild West herding cats instead of cattle. Kittens playing, older cats sunning themselves and barely looking up as cowboys thunder past, lassos raised and consistently, comically unsuccessful. The result was okay. Cats merged into each other at points. Sizes shifted inconsistently between frames — fine in a rough cut, not fine in anything client-facing. This concept would work better as illustration or traditional animation than as an attempt at photorealism. Useful thing to learn about where the tool's strengths actually sit.

‍

‍

So, is it fit for purpose?

For the record: yes, I know about the Will Smith spaghetti video and how far things have come since. Based on what I've actually put it through this week, these tools have real distance left to travel before they can replace a designer's judgement, not just a designer's hands.

Where it's genuinely useful right now: a cheap, fast way to test whether a creative concept has legs before committing real production time to it. A rough alternative to stock footage when you need something a stock library doesn't have. A single 8-second clip on Veo's faster model runs somewhere around a dollar or two in raw generation cost. Even the higher-quality model comes in well under what you'd pay for a single licensed stock clip, which typically runs anywhere from around $10 up past $100 depending on resolution and library.

The catch: raw generation cost isn't the real cost. A usable clip usually takes several attempts to land. Every clip still needs an editor's eye to catch the drifting design or the merging cats before it goes anywhere near a client.

Why this matters for Snifferdog specifically

We exist to make advertising as simple for a small business as sending an email. Upload a brief, approve a plan, launch a campaign, done. Creative is the obvious next piece. If a tool exists that can generate genuinely usable, on-brand creative end to end, we'll build it in.

But "we could technically bolt this on" and "this is fit for purpose" are different questions. We're not shipping the first one dressed up as the second. Right now, based on real testing rather than the hype cycle: Veo is a genuinely useful tool in a designer's kit. It is not yet a replacement for one. We'll keep testing, and we'll tell you honestly when that changes.

‍