Choose a source portrait
Start with a clear face image that supplies the subject's identity, lighting, clothing, and overall visual appearance.
What is a LivePortrait—and what can it do?
What is a LivePortrait? It is an AI model that transfers facial motion, expression, and head movement from a driving video onto a source portrait, creating an animated result while preserving the source image's identity and visual style.
LivePortrait in one line
The essential idea is simple: one portrait provides the appearance, and one motion source provides the performance.
- DefinitionLivePortrait uses a source portrait and a driving video to create a new animated portrait.
- CapabilityThe model can transfer expressions, eye direction, head pose, and other visible motion cues.
- LimitLivePortrait changes visible movement, but it does not turn one photograph into a complete real-world video scene.
- UsersArtists, researchers, educators, and developers use the model for visual experiments and character-driven media.
How it works
LivePortrait separates appearance from motion, then recombines them in a short image-to-video generation process.
Provide a driving performance
A video or motion sequence supplies cues such as facial expression, eye movement, head rotation, and speaking-related motion.
Render the animated portrait
The model maps the driving motion onto the source portrait and produces frames that preserve the subject while following the performance.
Can and cannot do
Understanding the boundary between motion transfer and full video generation helps set accurate expectations.
01
What it can do
LivePortrait can animate a single portrait with motion taken from another performance. It is suited to expressive talking-head clips, subtle reactions, head turns, eye direction, smiles, and stylized portrait animation. The source portrait can be photographic, illustrated, or otherwise visual, as long as the face and important features are readable.
- Transfers visible facial and head motion
- Preserves much of the source identity and style
- Works for realistic and illustrated portraits
02
What it cannot guarantee
LivePortrait is not a general-purpose scene simulator. It does not automatically invent a full body, new camera choreography, accurate hand movement, or a consistent background from one still image. Results can also weaken when the face is hidden, heavily angled, poorly lit, extremely stylized, or unlike the driving performance.
- Does not create every body movement from a face image
- May struggle with occlusion, extreme angles, or unusual faces
- Needs suitable source and driving inputs for stable results
Who uses it
The model is useful wherever a still character needs a compact, repeatable performance without filming a new subject for every variation.
-
1
liveportrait online
Compare the practical considerations of using LivePortrait through an online interface.
-
2
liveportrait ai
See how the AI model fits into the broader landscape of portrait animation tools.
-
3
how to use live portrait ai
Follow a focused workflow for preparing inputs and producing an animated portrait.
From research model to creative workflow
The format developed from a research problem—controlling portrait motion—into a practical way to explore characters, expressions, and visual communication.
-
A still image fixes the subject
Traditional portrait media begins with a fixed photograph or illustration. The appearance is established, but the image has no natural way to respond or perform.
-
A driving sequence adds performance
Motion-transfer research introduced a useful separation: one input supplies who the subject is, while another supplies how the subject moves.
-
Identity and motion are balanced
Modern portrait animation models aim to preserve recognizable features while following the expression and pose found in the driving source.
-
Animated portraits become reusable assets
The same source portrait can support different performances, making the format useful for experiments, character studies, explainers, and visual prototypes.
A visual before-and-after
Source portrait
The appearance comes from the source portrait; the movement comes from the driving performance.
See the idea in action
A useful way to understand LivePortrait is to compare the two inputs separately: the source image answers “who is being shown?” and the driving sequence answers “how are they moving?” From there, the model combines identity and motion into an animated portrait that can be reviewed, refined, and reused.
- Start with a clear, front-facing source portrait
- Use a driving clip with visible, readable expressions
- Review the result for identity, motion, and visual consistency
LivePortrait FAQ
A LivePortrait is an animated portrait generated by applying motion from a driving video or sequence to a source image. The result keeps the source subject's appearance while adding facial expression and head movement.
It is best understood as a portrait animation and motion-transfer model. It starts with an image, uses a separate motion source, and produces a short animated result rather than inventing an entire scene from text alone.
The usual workflow needs a source portrait and a driving performance. A clear face image and a motion source with visible expressions generally give the model more useful information to transfer.
It can transfer visible cues such as facial expression, eye direction, head pose, and related portrait movement. The quality depends on how well the source image and driving performance match the intended result.
It is not designed to guarantee full-body animation, complex hand movement, unrestricted camera motion, or a complete new environment from one portrait. Occlusion, extreme poses, unusual viewpoints, and mismatched inputs can reduce consistency.