Prompt Workbench
AI Video Prompt Builder
Start from a brief, not a blank prompt box. Pick your video model, and the builder gives you a shot plan, a paste-ready text prompt and a structured JSON version, set to that model’s clip lengths and aspect ratios, all in your browser.
Free with Google sign-in
AI Brief Assist
Paste a rough idea, product note, or messy offer and let AI fill the brief fields before you tune the final prompt.
Free while we test it. Buy me a coffee if it saves you time.
How the AI video prompt builder works
A video prompt that only says what you want to see leaves everything else to chance: the lens, the lighting, the pacing, the sound. This builder fills those slots from a short brief, so those decisions are yours rather than the model’s. It works the same way for every model it supports, and it changes only what that model needs.
The builder itself needs no account and costs nothing to run, because it writes prompts rather than video. Everything happens in your browser. You take the result to the model you chose. (AI Brief Assist, which fills the brief from a rough idea, is optional and needs a sign-in.)
Four steps from brief to prompt
- Pick a model and a format. Choose the video model, then the kind of video you want (UGC ad, product orbit, dialogue scene, cinematic reveal). A format preloads a shot structure you can edit.
- Describe the brief. Say what is on screen, who it is for and what the viewer should take away. This becomes the subject and action of the prompt.
- Set the look and sound. Choose the camera, lighting, aspect ratio and audio, and add any negatives. Anything you leave blank is something the model will invent.
- Copy the result. Take the Shot Plan, the paste-ready Text prompt or the JSON version, and run it on the model.
What changes when you change the model
The structure of a good prompt is the same everywhere. What differs is the envelope around it, and that envelope changes what you should write.
- Clip length. A short clip fits fewer beats than a long one. Fewer seconds means a tighter hook and one clear action; more seconds lets you plan a second moment. The builder only offers lengths the model accepts.
- Aspect ratio. The ratio decides the framing you describe. A vertical clip wants a subject centred and close; a wide one has room for context. Models differ in which ratios they offer at all.
- Audio. Most current video models generate sound with the picture, so the prompt should say what is heard: any spoken line, the effects, the ambience. Leave it out and the model picks a soundtrack for you.
- Negatives. There is no universal negative field. Where a model has none, the builder writes the negatives into the prompt as plain text. Whether a given negative holds is something to test shot by shot.
Models in this builder
Figures come from each vendor’s own documentation, checked on .
| Model | Clip length | Aspect ratios | Audio | Open |
|---|---|---|---|---|
| Veo Veo 3.1, Veo 3.1 Fast, Veo 3.1 Lite | 4, 6 or 8 s | 16:9, 9:16 | Native audio | Veo Prompt Builder |
| Gemini Omni Gemini Omni 1.1 Flash | 3 to 10 s | 16:9, 9:16 | Native audio | Gemini Omni in the builder |
| Grok Imagine Grok Imagine Video 1.5 | 1 to 15 s | 16:9, 9:16, 1:1, 4:3, 3:4, 3:2, 2:3 | Native audio | Grok Imagine Prompt Builder |
| Kling Kling 3.0, Kling 3.0 Omni | 3 to 15 s | 16:9, 9:16, 1:1 | Native audio | Kling Prompt Builder |
| Seedance Seedance 2.5, Seedance 2.0, Seedance 2.0 Fast | 4 to 30 s | 16:9, 9:16, 1:1, 4:3, 3:4, 21:9 | Native audio | Seedance Prompt Builder |
More models are being added. This table lists only what the builder supports today.
Text or JSON
Text is the paste-ready prompt for an app or chat-style interface, written as one block of prose. JSON is the same prompt split into named fields, which suits a batch where the camera, lighting and negatives must stay identical while the subject changes, and it diffs cleanly once a prompt becomes something you maintain. Neither is better: JSON is the same instruction with structure you can hold steady, so the gain is repeatability, not quality. The JSON prompt format guide covers when it pays off.
Where to go next
Rather start from something already written? The Veo prompt library and the Gemini Omni prompts collect prompts by use case, each labelled with how it was verified.
To choose between the two, read Gemini Omni vs Veo 3.1: what changes in your prompts. For the reasoning behind the structure, see the Veo prompt formula.