Make a UGC ad from a script

UGC ads are made on UGC Ads: a creator-style video of someone talking to camera. Only the words are needed. A presenter, your product, a place, a video to follow and the voices are all optional, and anything you add can be named in the words.

Make an ad

1. Write the words. Say where it happens, what they do, and the lines they say in double quotes. Only the quoted lines are spoken; if nothing is quoted, everything you wrote is read as the script. Under the box, a count shows how many words they say and about how long that takes.
2. Add what you want in it. Each group above the words is optional, and each tile shows the tag you name it by:
   • Presenters: someone from your Cast, a picture from your Library, or your own character, in one picture or up to four. A video of them works too: a still is taken from it. An ad takes up to four presenters, named @Presenter1, @Presenter2 and so on.
   • Product: a product from Products brings its name and its facts. Up to three pictures of it in all, named @Product1, @Product2 and @Product3.
   • Place: up to three pictures of the room, a prop or a look, named @Image1 and on. Or say where in the words.
   • Look like: a video for the model to follow, its look and its pace but not the people in it, named @Video1.
   • Voices: optional, one for each presenter (see below).
3. Name them in the words. Type @ to pick from what's in the bar. Words you paste that already say @Presenter1 or @Product1 are matched to the bar as they are.
4. Pick the models. The Video chip lists the models in two groups: Speak the words with the picture (Seedance, Kling 3.0, Kling 4.0 and Kling 4.0 Flash), and Lip-sync a voice exactly, which moves one face to a voice. Each model's line says the longest take it makes and its price a second. The Picture chip picks the model that draws the opening picture from your presenters, product and place, or No picture first, where the video model takes your pictures as they are.
5. Set the shape, size and length. Shape and Size list only what the model makes. Length starts on Fit the words, the shortest length that holds what is said; pick a length instead to make one take of exactly that. A lip-synced clip lasts as long as its voice.
6. Pick the captions and the takes. Captions shows your words captioned on your presenter, and sets the style, its size and place, and the AI label: see Captions and the AI label. ×1 to ×4 sets how many takes one press makes, each its own video from the same inputs.
7. Press Generate, or ⌘ Return (Ctrl Enter on Windows) in the words. The ad appears first in My ads. The words clear for the next one; everything else in the bar stays.

If something stops Generate, the line under the bar says what, such as "Say who is in your picture, and tick that they agreed." A script too long for one take is never refused: it's made in parts, joined into one video. See Long scripts: how an ad is made in parts.

Voices

A voice is optional. A model that speaks the words makes the sound itself, saying the lines you quote, so every presenter starts on the model's own voice. To choose how someone sounds, press their tile under Voices. ▶ plays a voice before you pick it. To take a voice out, press × on its tile: they go back to the model's own voice.

• The model's own voice: the video model voices it from the words.
• Their own: a Cast member's own voice, where they have one.
• A voice from the list, which you can filter and search.
• Record it now: up to 30 seconds, in your browser.
• Upload a recording: WAV or MP3. You're asked whose voice it is.

Voices offers only what the video model takes:

• Seedance takes any of them.
• Kling 3.0 takes a Cast member's own voice, kept with their character, and voices everyone else itself.
• Kling 4.0 and Kling 4.0 Flash take no voice, so Voices isn't shown: they voice the words themselves.

A voice you picked on another model stays in the bar for when you switch back, and the line under the bar says it isn't used. A lip-sync model needs a voice other than the model's own; there, a presenter from Cast starts on their own voice. A recording has to be a length the model takes; if it isn't, the line under the bar says so.

Real people

A photo or video of a real person can be used only with their permission. Open the tile, choose A real person, type their full name and tick the statement; for an AI character, a drawing or a mascot, choose that no real person is in it. Generate stays off until every presenter, the Look like video and any uploaded recording has an answer. Seedance doesn't take a real person's face as a reference, so an ad with a real person needs a picture model to draw the opening picture. See Real people: faces, voices and consent.

Templates

The Templates tab has eight formats, such as Unboxing, Problem → solution and Gym mirror, each a script and a place. Use this puts its words in the bar, with your product's name filled in when you've picked one from Products. Your models, presenters and product stay as they are, and nothing runs until you press Generate.

What it costs

The price is beside Generate. Hover over it, or move to it with the Tab key, to see the sum: pictures, clips, voice and the finish. The line under the bar says what one press makes and what each part costs. Each take is charged its own opening picture, clip and finish, which cuts it together and adds the captions when you pick a style; a voice read for the ad is made once and heard in every take, and a recording you bring adds nothing for the voice.

You're never charged more than the price you pressed at. If the price has gone up since the page loaded, you're shown the new one and asked to press again. If your balance is short, Generate opens a window with the price and the plan or credits that cover it.

When it's made

While an ad is made, its tile in My ads shows the stage it has reached, naming the model: the opening picture, the clip, then the finish, shown as Captions. It's made on our servers, so you can close the page.

A finished ad plays where it sits, with sound: press ▶. Open it to play it larger, switch between takes, Download it, or press Use again to put its words, people, product and settings back in the bar. Delete removes it for good.

Under its takes, References shows what the ad was made from, named as your words name them: each presenter (@Presenter1), the product's pictures (@Product1), your other pictures (@Image1), the video to look like (@Video1) and any recording you brought (@Audio1). Press one to see or play it where the ad plays; Back to the ad, or Esc, brings the ad back. A voice from the voice library isn't listed: there's no file of yours to open.

A take that fails says why. The credits for the step that failed come back automatically, and the steps after it aren't charged: see My video failed. Did I lose my credits?