Live Portrait
Live Portrait: animate still faces with video
A tradução deste card ainda não está disponível; parte do conteúdo aparece em inglês.
Live Portrait is a video generation model that animates a single portrait photo using motion from a reference video. It captures subtle facial expressions, head movements, and eye blinks while preserving the identity and style of the original image.
Use Live Portrait when you want to add life to a static portrait—for example, making a historical photo nod, giving a character a subtle smile, or creating a short looping video for social media. It works best with clear, front-facing portraits and simple driving motions.
What makes Live Portrait distinctive is its ability to transfer motion without requiring any 3D modeling or manual rigging. The output retains the original photo's lighting, background, and texture, resulting in natural-looking animations that feel authentic.
- Animating family photos for memories
- Creating expressive character avatars
- Enhancing digital art with subtle motion
- Generating short video greetings
- Bringing historical portraits to life
Maintains the original face's appearance, lighting, and background.
Accurately replicates head movements and expressions from a driving video.
Requires only a single photo and a short reference video.
- Use a clear, well-lit portrait with the face centered and eyes visible.
- Choose a driving video with simple, smooth motions for best results.
- Avoid photos with heavy shadows or accessories covering the face.
- Experiment with different driving videos to vary the animation style.
- May struggle with extreme head rotations or large facial movements.
- Works best with high-quality, front-facing portraits without obstructions.
- Generated video length is limited to a few seconds.
A vintage black-and-white photo of a woman, animated with a gentle smile and eye movement.
An elf warrior portrait from a fantasy illustration, nodding and looking around with subtle expression changes.
A close-up photo of a dog's face, animated with ear twitches and slight head tilt.
A number that controls randomness; using the same seed produces the same result.
Length of the generated video in seconds, typically matching the driving video.
Shape of the output video, like square (1:1) or landscape (16:9).
How closely the output follows the driving video motion; higher values mean more faithful motion transfer.
Preço sob consulta
Os preços exibidos são o que você paga na Muitto. Sem surpresas.
What kind of photo works best?
A front-facing, well-lit portrait with the face clearly visible and no obstructions.
How long is the output video?
The generated video is typically a few seconds long, matching the length of the driving video.
Can I use any driving video?
Yes, but simple, smooth head movements and neutral expressions yield the most natural results.