I make short videos for small businesses. Not the glossy sort with a crew and a catering van, the other kind. A florist who needs something for Instagram by Friday. A bakery in Stockport showing off a new sourdough. A personal trainer who wants fifteen seconds of himself looking like he knows what he’s doing. For years the work had a shape I could set my watch by: shoot on Tuesday, realise on Wednesday I was missing a cutaway, lose Sunday to re-editing the same forty seconds for the third time.
The filming was never the hard part. The middle was. The client who says, “Can I see it a bit warmer? And maybe one without the voiceover, and a square one for TikTok.” All reasonable requests. Each one an afternoon. I learned to quote for two versions and quietly build in the third, because the third always came.
So when the AI video tools started appearing, I left them alone for a good while. The first ones were fairground tricks: seven seconds of a melting labrador, people whose teeth kept changing their minds. Good for a laugh, no use to a paying client. I decided it was a bubble and went back to my tripod.
What changed my mind wasn’t a slick demo. It was a slow Thursday and a hummus brand. They wanted a launch clip and had the budget of a sandwich. Normally I’d have said no. Instead I spent an hour with one of the newer generators, describing the shots I couldn’t afford to stage: a jar turning slowly in warm light, a hand tearing flatbread, steam I’d never have caught on the day. Rough, but close enough to show the client what I meant. She approved the direction in ten minutes instead of two meetings. Then I shot the hero moments for real and let the machine handle the bits nobody would ever inspect frame by frame.
That’s the honest version of what changed. Not “AI replaced my job.” It took the boring part off my plate and left me the part I’m actually paid for.
Where it earns its keep
The wins are unglamorous, which is why I trust them. Versioning, mostly. A client wants the same idea in three moods, and I can rough all three before lunch and let them argue with a picture instead of a paragraph. The florist from the start of this was a good example. She wanted her Valentine’s clip to feel, in her words, expensive but not cold. I roughed four versions before the kettle had boiled, sent them across, and she picked the one I would have argued against. That is usually how it goes, and it is fine, because now we were arguing about the right thing instead of a vague feeling. Concept mockups, too: showing a nervous shop owner what “cinematic but not pretentious” actually looks like, before anyone has spent a penny. And filler. The establishing shot, the abstract texture behind a title card, the two seconds of nothing that a video needs so it can breathe. I used to buy that from stock libraries and it always looked like everyone else’s, because it was.
I’ve settled into a small rotation of tools rather than marrying one. The one I reach for first at the moment is Seedance 2.5, mostly because it gets me a usable clip on the first or second go rather than the tenth, and because the results hold together long enough to sit next to real footage without shouting about themselves. That last bit matters more than any feature list. A tool that dazzles on its own and then falls apart the moment you drop it into a timeline is worse than useless. It’s a time sink in a nice coat.
Where it still falls over
I want to be straight about the limits, because the breathless takes do nobody any favours. Hands are still a lottery. Anything with readable text, a shop sign, a label, a price, will betray you sooner or later. Continuity across a longer sequence drifts, so a person’s jacket quietly changes shade between two shots, and once you have noticed it you cannot unnotice it. And there is a stubborn five per cent, the uncanny remainder, where a clip is technically flawless and somehow lifeless, like a room tidied by someone who has never lived in it. You feel it before you can name it, and clients feel it too, even the ones who could not tell you why they keep asking for another take.
So the real footage has not gone anywhere. The face of the business, the actual product in the actual hands of the owner, the moment that has to feel true, that I still shoot on a Tuesday, still missing a cutaway, same as ever. What has gone is the dread around it. I no longer price a job assuming three lost weekends. I no longer turn down the hummus brand.
What I would tell someone starting out
Don’t lead with the technology. Nobody outside our little corner of the internet cares which model made the steam. Lead with the story, use the tools to take away the friction, and keep a human hand on anything that carries the brand’s real face. Learn to write a decent prompt the way you would learn to light a scene. It is a craft, it rewards being specific, and the gap between a lazy request and a considered one is the gap between something you bin and something you keep.
Mostly, though, I would tell them the thing I had to learn slowly, and it has nothing to do with kit. The expensive camera and the clever software were never really the point of any of it. The point was working out what a client needed before they could put it into words, and then getting out of the way of it. The tools got faster this year. The job, the real one, is sitting exactly where it always did. I just get more of my Sundays back.
