Skip to main content

Generating Videos

Generate AI video takes for each scene in your storyboard using the AI Director chat.

Opening the video generation dialog​

  1. Click on any scene in your storyboard
  2. Open the video generation dialog for that scene
  3. The AI Director chat interface appears, where you can configure and generate takes

Choosing a generation type​

Select the type of generation that fits your needs:

TypeDescriptionSource input
Image to videoAnimate a still image into a videoStart frame, optional end frame
Video to videoEdit or transform an existing videoVideo source

Source inputs​

Depending on the generation type, you can provide:

  • Start frame — The first frame image for image-to-video generation
  • End frame — The target last frame (supported by some models)
  • Video source — An existing video to edit or transform
  • References — Additional images or videos for style guidance

Writing your prompt​

Describe what you want the video to show. Each model has a character limit displayed in the input area.

Multi-shot prompts​

Some models support multi-shot generation, where you define multiple prompts for different segments of the video:

  1. Enable multi-shot mode in the generation input
  2. Add individual shots, each with its own prompt and duration
  3. The total video duration is the sum of all shot durations
  4. Each shot can reference different assets with @ mentions

Multi-shot is useful for creating videos with distinct movements or scene transitions within a single generation.

Configuring generation settings​

Model and duration​

Use the model selector to pick which AI video model to use. Each model offers different durations — select the one that fits your scene.

Aspect ratio​

Set the aspect ratio for the generated video. Available ratios depend on the selected model and include options like 1:1, 3:4, 4:3, 9:16, 16:9, and 21:9.

Audio generation​

Toggle audio generation on or off. When enabled, the AI generates a matching audio track alongside the video.

Prompt enhancement​

Like image generation, you can choose between:

  • Fine-tune — Your prompt is enhanced for better results
  • Raw — Your prompt is sent exactly as written

Generating a take​

  1. Configure your generation type, source inputs, prompt, and settings
  2. Review the credit cost shown on the generate button
  3. Click generate to submit your request
  4. The take starts generating and appears in the chat history

Viewing and managing takes​

  • Generated takes appear in the AI Director chat as video messages
  • Click any take to preview it in the scene player
  • The most recent take is selected by default
  • You can generate as many takes as you want — try different prompts, models, or durations to explore creative directions

Downloading takes​

Single take​

With a take selected, click Download to save the video file.

All takes​

Click Download All in the action bar to download all ready videos at once.

Tips for better results​

Write clear prompts​

Be specific about what you want:

  • Describe the action, mood, and visual style
  • Include details about camera movement if relevant
  • Reference your narrative and visual descriptions

Provide good source frames​

For image-to-video generation, the quality of your start frame matters:

  • Use high-quality images that represent the first moment of your shot
  • Include an end frame to guide where the video should go

Iterate quickly​

  • Start with a simple prompt
  • Generate a take to see results
  • Refine your prompt based on what you see
  • Try different models — each produces different styles

Use multi-shot for complex scenes​

When your scene has multiple beats or camera movements, multi-shot gives you finer control over each segment.

Troubleshooting​

Generation is taking too long​

Video generation can take a few minutes. The status indicator shows progress.

Take failed​

If a take fails:

  • Check your prompt for any issues
  • Try a different model
  • Adjust the duration
  • Generate a new take

Video doesn't match expectations​

  • Refine your prompt with more specific details
  • Provide better source frames
  • Try a different AI model
  • Use multi-shot to break down complex movements

Learn more​