Getting Started with AI Video Generation

What is AI Video Generation?

AI video generation uses machine learning models to create video clips from text prompts, images, or a combination of both. Instead of filming with a camera or animating frame-by-frame, you describe what you want and a model like Kling 3.0, Veo 3.1, or Sora 2 Pro produces a video clip in seconds.

Prism gives you access to all of these models in one workspace so you can pick the best tool for each shot.

Your First Video in 3 Steps

1. Write a prompt

Start simple. A good beginner prompt describes the subject, action, and style:

A golden retriever running through a sunflower field at sunset, cinematic slow motion, shallow depth of field

Avoid vague prompts like "a cool video." The more specific you are about lighting, camera angle, and mood, the better the output.

2. Choose a model and settings

Each model has strengths:

• Kling 3.0 — Great for realistic motion and human subjects. Supports audio generation.
• Veo 3.1 — Google's model with 4K output and built-in audio. Excellent for cinematic shots.
• Sora 2 Pro — Strong at creative and stylized content.
• Wan 2.1 — Fast and cost-effective for iteration.

For your first video, start with a 5-second clip in 16:9 aspect ratio. You can always generate longer clips once you're happy with the style.

3. Generate and iterate

Hit generate and wait for the result. If the output isn't quite right, tweak your prompt rather than starting from scratch. Small changes like adding "handheld camera" or "overcast lighting" can dramatically shift the result.

Text-to-Video vs. Image-to-Video

Text-to-video (T2V) generates everything from your text description. It's the simplest way to start but gives the model full creative control over composition.

Image-to-video (I2V) uses a reference image as the first frame, then animates it based on your prompt. This gives you much more control over the visual style and composition. Generate an image first in the Image tab, then send it to video generation.

Tips for Better Results

• Be specific about camera movement. Prism supports 17 camera movements — dolly, orbit, pan, tilt, drone shot, and more. Adding "slow dolly forward" to your prompt creates intentional, cinematic motion.
• Specify the mood. Words like "warm," "moody," "high contrast," or "pastel" guide the model's color grading.
• Use reference images. Upload style references to keep a consistent look across multiple clips.
• Start short. Generate 5-second clips first. Once you find a style you like, extend to 10 seconds.

What's Next?

Once you're comfortable generating individual clips, try the Storyboard to plan multi-scene videos, or use the Timeline Editor to arrange clips into a complete video with captions and transitions.