How does the AI ad prompt builder work?
The builder is a form that writes the prompt for you in the right order. You pick a tool and an ad format, fill in the fields, and the prompt appears in the panel as you type, ready to copy.
- Pick a tool. Seedance 2.0, Kling, Veo 3.1, Higgsfield or Nano Banana. The form shows only the fields that tool's guide uses.
- Pick an ad format. UGC selfie, podcast clip, street interview, short drama, claymation, animation, product shot or brand film. The format becomes the opening line of the prompt.
- Fill in the fields. Subject, action, setting, camera, motion, lighting, style, dialogue and sound for video tools; product, composition and on-image text for Nano Banana.
- Add duration and aspect ratio. The builder puts them in the prompt only when the tool's guide does. Otherwise it lists them under the prompt as settings to choose in the tool.
- Copy the prompt and paste it into the tool.
Where does the order of each prompt come from?
Each tool's order is taken from its official prompt guide and stored with the guide's link and the date it was checked, 12 Oct 2026. The builder shows that link under every prompt.
| Tool | Documented order the builder follows | Dialogue syntax in the guide | Source |
|---|---|---|---|
| Seedance 2.0 | Subject, action, scene, lighting, camera, style, then audio | Curly braces | BytePlus prompt guide |
| Kling | Subject, action, setting, camera, lighting, then sound | Speaker (tone): line | Kling 4.0 guide, VIDEO 3.0 guide |
| Veo 3.1 | Camera, subject, action, context, style and ambiance, then audio | Quotation marks | Google Cloud Veo 3.1 guide, Gemini API docs |
| Higgsfield | Subject, action, setting, camera, then light, style and sound | Speaker: "line" | Higgsfield help center |
| Nano Banana | Subject, action, location, composition, style, with text in quotes | Not a video model | Google Cloud Nano Banana guide |
The builder writes spoken lines in quotation marks for every video tool. For Seedance, change them to curly braces, and for Kling, put the speaker and tone before the line, as the guides above show.
Worked example: a Veo 3 UGC selfie prompt
Enter these values in the builder above to see how it orders a Veo prompt:
| Field | Value |
|---|---|
| Tool | Veo 3 (listed as Veo 3.1 in the menu) |
| Ad format | UGC selfie |
| Subject | a woman in her 30s in a kitchen holding a steel water bottle |
| Setting | a bright apartment kitchen in the morning |
| Camera | handheld selfie, front camera, medium close-up |
| Motion | she turns the bottle to show the label |
| Dialogue | This thing keeps ice for a whole day. |
| Duration | 8 seconds |
| Aspect ratio | 9:16 |
The builder output appears above as soon as these are entered. Its structure follows Google Cloud's formula, which starts with the shot:
- An opening line naming the format, a UGC selfie ad.
- The camera line first: handheld selfie, front camera, medium close-up.
- The subject: the woman in her 30s holding the steel water bottle.
- The motion: she turns the bottle to show the label.
- The setting: the bright apartment kitchen in the morning.
- The spoken line in quotation marks after who says it, as the Gemini API guide asks for speech. Fill in "Who speaks" to name the speaker.
- Under the prompt, not inside it: duration 8 seconds and aspect ratio 9:16, to set in Veo's controls.
Two things to add by hand. First, a sound line: Google's guide describes sound effects and ambient noise in their own sentences, for example a fridge hum and a bottle cap twist, and the builder's sound field is there for that. Second, keep to one short line per clip, since 8 seconds is the longest single Veo generation. Veo clips with reference images must be 8 seconds, per the Gemini API docs.
For more Veo templates by format, see Veo 3 prompts for ads.
How should you use the prompt once it is built?
Treat the output as the first draft of one shot, not a finished ad. An ad is usually several shots: the hook, the product moment and the call to action, each generated and cut together.
- One beat per prompt. Put the hook line in one clip and the call to action in another.
- Change one field at a time. If the first result is wrong, edit the field that caused it and regenerate. Higgsfield's help center gives the same advice.
- Keep the presenter fixed. Use the same subject wording and the same reference images for every shot of one ad.
- Swap tools without rewriting. Keep your entries, switch the tool, and the builder reorders them for the new guide.
The prompt library has full templates for each format, with placeholders for the parts the form doesn't cover, such as multi-shot storyboards.
What are the limits of the prompt builder?
The builder orders and labels your words; it doesn't judge them. It can't tell whether a shot will generate well, and it has no outputs to show, because it never calls a model.
- No guarantee of results. The order comes from each tool's guide, but every model still varies between generations.
- One shot per prompt. Multi-shot syntax (Seedance's Shot 1, Shot 2, or Veo's timestamp prompting) is on the tool pages, not in the form.
- Dialogue follows each guide. The builder writes the spoken line in the tool's documented form: curly braces for Seedance, a speaker label and colon for Kling, quotation marks after the speaker for Veo, and a speaker label with a quoted line for Higgsfield. Check the guide if you use more than one speaker.
- Guides change. Models update often. Each prompt links to the guide it follows and the date it was checked.
- Claims are your responsibility. The builder does not check the line against ad platform rules. Make only claims the brand can legally make.
- Nothing is stored. What you type stays in your browser tab and is cleared when you close it.
If you make AI ads that already perform on paid traffic, see how Anglr works for creators. All free tools are listed on the tools page.
FAQ
Is this AI video prompt generator free?
Yes. It needs no sign-up and no credits, because it fills a template instead of calling an AI model. What you type stays in your browser tab.
Does the prompt builder use AI to write the prompt?
No. It puts your fields in the order each tool's official prompt guide documents, adds the labels that guide uses, and leaves the creative choices to you. The same inputs always give the same prompt.
Why are duration and aspect ratio not in my prompt?
Seedance, Kling, Veo and Higgsfield set them as settings or API parameters, not prompt text, so the builder lists them under the prompt as settings to choose. Nano Banana's own templates end with the aspect ratio, so for Nano Banana it goes in the prompt.
Which tool should I pick for a talking UGC ad?
Any of the video tools can take a spoken line. Veo and Seedance generate the voice with the clip, and Kling ties each line to a named speaker. Compare the syntax on the prompt library pages before you choose.
Can I write a Seedance 2.0 prompt with this?
Yes. Choose Seedance 2.0 and the fields follow BytePlus's order: subject, action, scene, lighting, camera, style, then dialogue and sound. BytePlus writes dialogue in curly braces, and the builder writes the line that way for Seedance.
Does the builder save or send my prompt?
No. There is no server call and no account. Copy the prompt with the button, and it is gone when you close the tab.
Sources
- BytePlus ModelArk – Dreamina Seedance 2.0 series prompt guide
- Kling AI – Inside Kling 4.0: A Complete Guide to AI Video Creation
- Kling AI – Kling VIDEO 3.0 Model User Guide
- Google Cloud – The ultimate prompting guide for Veo 3.1
- Google AI for Developers – Generate videos with Veo (Gemini API)
- Higgsfield Help Center – How to Write a Good AI Prompt
- Google Cloud – The ultimate Nano Banana prompting guide
Checked: 12 Oct 2026