Ad generation

Ads made from your brand, not a template.

Statics, video and UGC on your real product shots, with the headline set in your type. Every image and video model on one subscription, and the agent writes each prompt the way that model works best.

Hero ad made by SandstormHero
Lifestyle ad made by SandstormLifestyle
Flat lay ad made by SandstormFlat lay

What it makes

Ask in plain words. It comes back with finished ads, in every size you run.

StaticsHero, lifestyle, flat lay, native-UI mockups. Literal copy set in your real font, your logo placed. Exported in 4:5, 1:1 and 9:16.
Video adsMulti-scene clips with one look carried across every scene. Voiceover, captions and music, stitched into 9:16, 4:5 and 1:1 masters.
UGC and talking headsCreator-style testimonials and talking heads with native audio, lipsync and a cloned or library voice.
Product shotsPackshots on clean or styled backgrounds. Background removal, expand and upscale on the shots you already have.
Fixes and variationsA wrong price, a typo, a copy swap: it fixes the one element and leaves the rest alone. Or it makes ten more like the winner.
Editable layersLayer stacks you can rework, and send to Figma as real layers rather than a flat image.

How an ad gets made

Four steps, none of them yours.

01It starts from your brandProducts, offers, voice, type, logos and product shots, learned from your site and your accounts. Never a template.
02One concept per adA strategist brief for each ad: the angle, who it is for, the headline. Different angles, not the same ad three times.
03One prompt per modelEvery model wants something different. It writes each prompt the way that model works best, and picks the model per shot.
04Render, type, sizesThe headline set in your type, the logo placed, every placement size exported and saved to your library.

One ask, three statics

It pulls the product shots, colours and voice it already has, then writes one full prompt per ad.

You say
“Make three static ads for the Marsh Cap. Different angles.”
It pulls the product shots, colours and voice it already has for Goodwater, then writes one full prompt per ad, each for the model that does that shot best.
Sandstorm writes · one prompt per ad
GPT Image 2.5 logoGPT Image 2.5Product hero · 4:5Editorial product photograph of the Goodwater Marsh Cap: waxed-canvas six-panel cap in tan (#C9B79C), dark brown leather strap, embroidered pointer on the front panel exactly as in the reference image. Cap rests on weathered grey driftwood; marsh grass soft and out of focus behind. Overcast dawn light from camera left, soft shadows, light moisture beading on the brim. 35mm lens, f/2.8, eye level, slight downward tilt. Palette: tan, slate, muted green. Keep the top third clear for a headline. No text, no hands, no additional logos.
You get
Hero ad made by SandstormHero
Lifestyle ad made by SandstormLifestyle
Flat lay ad made by SandstormFlat lay
Three ads, each exported in 4:5, 1:1 and 9:16, with your headline set in your type.

One ask, one 20-second ad

Four scenes, one look. Each scene references the last frame of the one before.

You say
“Cut a 20-second ad. The dog, the marsh, the cap. Make it feel like one morning.”
Sandstorm writes · four scenes, one look
Seedance 2.5 logoSeedance 2.5Scene 1 · Dawn · 5sHandheld follow through marsh grass at blue hour, black lab runs ahead and out of frame, 35mm, low contrast, cool palette, ambient marsh audio, no music yet.
Same dog, lens, light and colour carried across all four; each scene references the last frame of the one before.
You get
DawnTurnCapHome
0:000:20
VoiceoverCaptionsMusic9:16 · 4:5 · 1:1
One 20-second ad, stitched, in every size you run

Every model, and it picks

Prompting these models well is hard, and every model wants something different. It uses the right technique for every shot.

One subscription, every modelIt picks the right one for each shot. No stacking six tools
FLUX logoFLUXGemini logoGeminiGPT-Image logoGPT-ImageGrok Imagine logoGrok ImagineHedra logoHedraHiggsfield logoHiggsfieldHunyuan logoHunyuanKling logoKlingMiniMax Hailuo logoMiniMax HailuoNano Banana logoNano BananaRunway Gen-4 logoRunway Gen-4Seedance logoSeedanceSeedream logoSeedreamSora logoSoraVEED logoVEEDVeo logoVeoWan logoWanand more

Questions

Do I need my own product photos?

It works from the product shots already on your site and in your ad accounts, and you can upload more any time. Real product shots, not stock.

Which sizes does it export?

Statics in 4:5, 1:1 and 9:16 by default, with the headline re-set for each. Video masters in 9:16, 4:5 and 1:1. Ask for another size and it makes that too.

Can I change one thing without redoing the ad?

Yes. Tell it what to change. It fixes that one element and leaves the rest alone, and layer stacks open in Figma as real layers.

Does it use my logo and my font?

Yes. It sets the copy in your brand's real font and places your brand mark from your brand identity, sized for the format.

Which models does it use?

Every image and video model on one subscription: GPT Image, Nano Banana, Gemini, Seedream, FLUX for stills; Seedance, Veo, Sora, Kling, Hailuo, Wan and more for video. It picks per shot; you never pick a model.

Where do the ads go?

Into your asset library as permanent links, and straight into your ad accounts when you say publish.

Your first ads are minutes away.

Connect the brand, ask for three statics, and see what it does with your own product shots.

For one marketer or the whole team. Cancel any time.

More from Sandstorm