Studio V2.7 Model

We're excited to announce the release of our Studio V2.7 Model, now available in Beta for paying users!

V2.5 introduced our most advanced model for image-to-video talking head generation, with support across audio domains (talking, singing, laughing, and more) and image styles (cartoons, photorealistic, stylized, and more). V2.7 goes further — adding full-body movement and background motion to make your videos even better.

What’s New in V2.7

• Full-body movement - videos generated with V2.7 will include context-aware upper body movement and gestures that make your videos more natural and less "robotic". Is your rock star holding a guitar? Now they can play it!
• Background motion - elements in your background that were previously "frozen" with V2.5 will now be fully animated. Flames will flicker, smoke will rise, and flowers will move in the wind.

Input Examples

Here are some examples that work well with V2.7. All videos were generated from a single image and audio prompt.

Examples of what to avoid with V2.7

❌ Close-up portraits: If the image is a relatively narrow face crop and only includes the face or shoulders, use V2.5 instead.

❌ Multiple people: V2.7 works best on images with only a single person in the frame. Outputs with more than one person may fail to produce desired results.

❌ Animal inputs: V2.7 currently works only if a human face can be detected in the image. Use V2.5 instead for animals.

❌ Hands around the face: V2.7 works best on images where the hands are far away from the face to start.