AI Video turns a guest's photo into a short cinematic clip. You set the look once in a template, and every guest gets their own version.

AI Video events need a **Premium** licence to run. You can build and test one on any plan — the licence is checked when the event goes live.

### **Build the template**

1. Go to **Events** and click **Create New Event**.

2. Name your event and finish creating it.

3. Open the **Flow** tab, open **Choose Experience**, and click **Add template**.

4. Pick your canvas size. **1x1**, **16x9** and **9x16** are the usual choices for video.

AI outputs support ten aspect ratios: **1:1, 16:9, 9:16, 21:9, 3:2, 2:3, 5:4, 4:5, 4:3, 3:4**. The editor tells you if a size is not supported.

### **Add the effect**

1. Add a **Capture** element and set its Output Type to **Image**.

2. With the capture selected, open its **Settings**, then **Effect** → **Add effect**.

3. Choose **AI Video** and pick your prompt.

4. Set **Quality** — **Low**, **Standard** or **High**.

Quality changes the token cost, and that cost multiplies by every guest. Token prices are set for your account, so check your own **Billing** page before a large event rather than relying on a figure from an article.

If your prompt uses a reference image added with **@**, keep that reference in the **image prompt** only, not the video prompt.

If your prompt does not describe camera movement, Pictor applies a slow cinematic motion by default. Override it in the prompt if you want something else.

![AI Video, added to the capture element.](https://app.chatwoot.com/rails/active_storage/blobs/redirect/eyJfcmFpbHMiOnsibWVzc2FnZSI6IkJBaHBCSkx5d1FZPSIsImV4cCI6bnVsbCwicHVyIjoiYmxvYl9pZCJ9fQ==--bfbb41c3ac17d651dc3de6872e6580505c60f57d/91-ai-video-effect.png?cw_image_width=auto)

*AI Video, added to the capture element.*

### **Position it**

Place the AI output where you want it on the canvas and use **Layers** to control what sits in front of what. Make the timeline at least as long as the video.

### **Test it**

1. Open the **Launchpad** and click **Simulation test**.

2. Take a photo and answer any prompt questions.

3. Wait for the render.

| **Output** | **Typical render time** |
| ---------- | ----------------------- |
| AI stills  | around 37 seconds       |
| AI videos  | 1–2 minutes each        |

Check the result under **Sessions** → **View Output**.

Budget that render time into your event plan. At 1–2 minutes a clip a queue builds quickly — see **On Screen Delivery** for how to keep guests occupied while they wait.

### **Multi-scene videos**

You can stitch several AI clips into one video.

1. Create a new template.

2. Add **Capture 1** with an **AI Video Output**, then **Capture 2** with its own **AI Video Output**.

3. Set the timeline long enough for both.

4. Hide the captures so only the AI results are visible.

5. Arrange the clips side by side on the timeline.

The results stitch into one continuous video.

### **Overlays and blend modes**

Video overlays add transitions and texture — white flashes, strobes, light leaks.

1. Add a **Video Asset** to the template.

2. Resize it to match the canvas.

3. Add a **Blend Mode** and choose **Screen**.

In Screen mode dark areas become invisible and bright areas stay, so a white-flash clip reads as a transition rather than a black box.

Adding a video element to a template is itself a **Premium** feature. On an AI Video event that changes nothing, but bear it in mind if you reuse the technique on a simpler template.

### **Rendering happens in parallel**

Multiple scenes render at the same time, not one after another. Two one-minute videos take roughly one minute plus stitching, not two.

### **Reviewing the result**

Go to **Sessions** and click **View Output**. You will see each scene, the overlays, and the combined video.

### **See also**

* **Custom AI Prompts** — writing the prompt behind it

* **Plans and features** — AI video needs Premium

* **Pricing plans** — what tokens cost