Ads for products, generated
Paste a product. Hazevio studies what is winning in its niche, then renders presenter videos, product-only videos and static ads that read as filmed, and publishes them where you sell.
Keep scrolling. That is a Hazevio ad, and you are moving through it a frame at a time.
One product photo went in. The presenter was cast for the brand, the setting chosen from what is winning in skincare, the script written to say nothing a customer would have to have said.
And this is the caption it ships with.“Dramatization. Presenter is AI-generated, not a real customer.”
Three formats from one product
One product photo went in. These came out of the same pipeline, for a brand invented for this page. The presenter is one strength, not the whole story.
An original on-camera presenter, cast for your brand.
Candidates are generated from your brand profile; you pick one. They hold your actual product, in a setting and pace taken from what is currently winning in your niche, and say a script that never claims to be a customer.
Ships to TikTok, Reels, Shorts, Facebook, Threads, LinkedIn
The product, filling the frame. Nobody in it.
Flat-lay and in-hand video for the surfaces where a talking head loses: listing pages, muted feeds, catalogue tiles. Slow push-ins, real light, the finish and texture doing the selling.
Ships to Etsy listings, Reels, TikTok, Shorts
Stills for every placement a seller actually buys.
The same angles as the video, rendered as photographs and laid out for eleven placements: square and portrait feed, story, pin, listing image, four display sizes and an email hero. Nothing baked into the picture; every word stays editable.
Ships to Meta feed and story, TikTok, Pinterest, Etsy, Google Display, email
Three brands were invented for this site, in three niches, and put through the same pipeline. Presenters shown are AI-generated.
How an ad gets made
Each one consumes the one before it. Every paid stage claims its place in the sequence before it spends, so a job that runs twice bills once. The phone shows what each stage produced for the ad above.
Your Etsy shop, your Shopify store, or one product URL. Nothing runs until there is artwork to work from: an imageless product would pay for two stages before the missing asset was caught.
On the phone: your product artwork
A vision pass over what is currently performing in that niche, reduced to a structured blueprint: framing, lighting, energy, beat timing. The format, not the footage. The blueprint has no column for a frame or a transcript.
On the phone: a real blueprint
For a presenter ad, candidates are generated from your brand profile and you pick one. Casting is text-to-image only; nothing is composited from anyone's likeness. Product-only ads skip this step, and its cost, entirely.
On the phone: candidates, you choose
Your actual product is put into the frame before any video is generated, so the thing in the ad is your thing. Then a script the presenter can say without claiming to be a customer, or no script at all, and a 9:16 render.
On the phone: the composite
Checked for black frames, frozen picture and dead audio before anyone sees it. A silent render is recorded as failed, never as ready. The disclosure is already in the caption.
On the phone: ready. caption attached.
Two things it will not do
It will not reuse a real person. It will not hand over an ad without the disclosure.
Most of this category will happily rebuild a real creator's video and let you post it as a testimonial. Both of those are someone else's problem right up until they are yours.
The obvious way to build this is to take a frame of a creator who is performing well and regenerate them. That is a modified likeness of an identifiable person, used to sell something.
So what gets extracted is the format — framing, lighting, room, energy, beat timing — and any presenter is synthesized from your brand instead. The blueprint has no column for a frame or a transcript, scripts are checked against the source and rejected past seven consecutive shared words, and the compositor refuses any reference image that is not on our own origin.
A synthetic presenter who reads as a satisfied customer can be treated as a fake testimonial under the FTC Endorsement Guides and the 2024 Rule on Consumer Reviews and Testimonials.
So the disclosure is enforced in code, at the boundary where an ad leaves the system, not merely suggested where the caption is written. The script prompt forbids first-person experience claims, and the output is checked afterwards. A prompt is a request; a check is a control.
Where it goes
You connect the accounts. Hazevio posts the ad, reads the numbers back, answers what it stirs up, and lets you put money behind the one that worked.
Views, likes, comments and shares per post, pulled back from each platform and rolled up on a calendar, so you can see which ad earned the boost.
Comments and DMs the ad brings in are answered from a script you approve, on Instagram and Facebook, through the same first-person-claim gate as the scripts.
The ones that work can be pushed into paid campaigns on your own ad account. Spend stays on your account, never on ours.
Everything in the box
Each line below is already on this page somewhere. This is the list, marked off as you read it.
An original on-camera presenter, cast from your brand profile, holding your actual product.
Flat-lay and in-hand, nobody in it, for listing pages and muted feeds.
The same angles as photographs, laid out for eleven placements. Every word stays editable.
What is winning in your niche, reduced to a blueprint of framing, lighting, energy and beat timing.
An Etsy shop by name, a Shopify store you connect, or one product URL. Prices keep their currency.
The real product goes into the frame before any video is generated, so the thing in the ad is your thing.
Black frames, frozen picture and dead audio are caught before you see the ad. A silent render is recorded as failed.
Every ad leaves with the AI disclosure in its caption. Publishing refuses a payload without it.
Scripts are rejected past seven consecutive words shared with the source. No frame, no transcript, no borrowed likeness.
Each paid stage shows its estimated cost before it runs. A job that runs twice bills once.
TikTok, Instagram Reels, Facebook, Threads, YouTube Shorts and LinkedIn, from accounts you connect.
Views, likes, comments and shares per post, pulled back from each platform onto a calendar.
Comments and DMs the ad brings in, answered from a script you approve, on Instagram and Facebook.
The ones that work, pushed into paid campaigns on your own ad account. Spend stays on your account.
From sellers
Founders, owners and marketing leads on running their products through Hazevio.
One listing in, and the presenter cut, the product-only cut and the statics all came from the same product. The estimate showed up before anything rendered, and that is the part I kept checking.
Nothing spends until we approve the estimate. That one step is why I was comfortable pointing it at our own shop.
The script got sent back once for repeating a line from the video it studied. I would rather it be strict than have us sound like everyone else.
The disclosure is already in the caption by the time a post reaches us. Nobody here has to remember to add it.
A render that failed its own check never reached us. The one that passed is the one we posted.
It took our shop by name and had the products, the prices and the currency right. Nobody here typed a brief.
Our bottle stayed our bottle. The presenter was built around the real product, not something that looked like it.
The statics came back sized for every placement we use, feed to email, from one pass.
The product-only cut does the job when we do not want a presenter at all. Same product, no face.
Questions
Three formats from one product: a presenter video with an original on-camera presenter, a product-only video with nobody in it, and static ads laid out for eleven placements. All three are checked and captioned before you see them.
No. Presenters are generated from your brand profile with text-to-image only. Nothing is composited from anyone's likeness, and the pipeline refuses any reference image it did not create itself.
No. Product-only video and static ads have nobody in them. For listing pages and muted feeds they often convert better, and they skip the casting stage and its cost.
Yes. Every ad leaves Hazevio with the disclosure written into its caption, and the publish step refuses a payload without it. It is a control in code, not a suggestion in a prompt.
No. It extracts the format — framing, lighting, energy, beat timing — into a blueprint that has no column for a frame or a transcript. Scripts are checked against the source and rejected past seven consecutive shared words.
TikTok, Instagram Reels, Facebook, Threads, YouTube Shorts and LinkedIn, from accounts you connect. Comment and DM replies run on Instagram and Facebook.
An Etsy shop, imported by name, or a Shopify store you connect. You can also start from a single product URL. Prices are kept only with their currency, so an ad never quotes a figure in the wrong one.
They were rendered by the same pipeline you would use, for three brands invented for this site: Verdant Lab, Ferro & Oak and Kite Audio. Presenters shown are AI-generated.
Hazevio is not open for general signup yet. Each paid stage shows its estimated cost before it runs, and a job that runs twice bills once. Request access and say what you sell.
Hazevio is running against real products and real spend, and is not taking general signups. If you sell something and want it in the next group, say what you sell and where you sell it.
