Reference Images vs Reference Videos: When to Use Each

Larpable uses two types of reference inputs: reference images (for photo generation) and reference videos (for video generation). Understanding what each does and when to use them is key to getting the results you want.

Reference Images

A reference image — also called a scene image — is a photo that defines the environment, lighting, composition, and outfit for a generated photo. When you upload a scene image, Larpable identifies the person in it and replaces them with your AI identity.

Reference images are used for photo generation. They determine where your identity appears, what they're wearing, and how the photo is composed.

When to Use Reference Images

• When you want a static photo of your AI identity in a specific setting
• When you need content for Instagram posts, Pinterest, or blog images
• When you want to control the exact environment, lighting, and composition
• When you want to create a base image that you'll later turn into a video

Reference Videos

A reference video is a short clip that defines the motion for a generated video. Larpable takes the visual appearance from an image and the motion pattern from the reference video, then combines them to create a video of your AI identity performing that motion.

Reference videos are used for video generation. They determine how your identity moves — turning, walking, gesturing, smiling.

When to Use Reference Videos

• When you want dynamic, moving content for social media
• When you need video for TikTok, Reels, or YouTube Shorts
• When you want your AI identity to perform a specific action or motion
• When you have a good static image and want to bring it to life

How They Work Together

In a typical workflow, you use both: first, you generate a photo using a reference image, then you turn that photo into a video using a reference video. This gives you control over both the appearance (via the image) and the motion (via the reference video).

On Larpable's canvas, you can connect an image generation node's output to a video generation node's image input, creating a pipeline from scene to photo to video.

Quality Tips

• Use high-quality, well-lit reference images for the best photo results
• Use smooth, short reference videos with natural motion for the best video results
• Match the framing between your image and reference video for the most natural video results
• Build libraries of both types so you always have good references on hand

Next Steps

Read our guide to AI influencer videos and guide to realistic AI photos for detailed instructions on using each type of reference.