GPT Image 2.5 API is live. Create and edit images →
PoYoPoYo
WAN ANIMATE 2 · COMING SOON

High-fidelity motion transfer, identity preservation, and text-driven camera control

Wan Animate 2 API: High-Fidelity Character Animation and Motion Transfer

Turn a character image into a complete performance driven by a reference video. Wan Animate 2 transfers frame-aligned body movement, facial expressions, hand motion, and fine performance details while preserving the reference character's appearance and body structure.

Text-driven viewpoint control separates the generated camera angle from the driving footage, giving creators a direct way to reinterpret one performance across different character styles and camera positions.

Core Features of Wan Animate 2 API

Direct video conditioning, identity preservation, and controllable viewpoints turn a single recorded performance into a flexible character animation source.

01

Wan Animate 2 API for End-to-End High-Fidelity Character Motion

Wan Animate 2 API takes a character reference image and a driving video, then injects frame-aligned motion conditions directly inside its Diffusion Transformer. The end-to-end character animation pipeline preserves facial expressions, hand movement, body poses, and performance timing without a separate skeleton or keypoint extractor, supporting high-fidelity motion transfer from subtle expressions to large full-body actions.

  • Animate a reference image from a driving video
  • Preserve frame-aligned facial, hand, and body motion
  • Remove intermediate skeleton and keypoint extraction

02

Wan Animate 2 API for Cross-Style Character Animation

Wan Animate 2 API animates people, cartoon characters, robots, and animals across different visual identities and body structures. When the reference character differs significantly from the driving performer in appearance, proportions, or shape, the model retargets the motion to the target body while preserving its defining design and identity, producing coherent cross-character animation.

  • Animate people, cartoons, robots, and animals
  • Retarget motion across different body shapes
  • Preserve character appearance, proportions, and identity

03

Wan Animate 2 API Text-Controlled Camera and Multi-View Animation

Wan Animate 2 API uses an optional Viewpoint LoRA to decouple the generated camera from the driving video's original angle. Natural-language prompts select among 12 azimuth orientations and 4 elevation angles for side, top, bottom, and eye-level views, turning one motion reference into consistent multi-view character animation.

  • Control character animation viewpoints with text
  • Choose 12 azimuth and 4 elevation angles
  • Generate multiple camera views from one performance

04

Wan Animate 2 API Identity-Preserving Motion Transfer

Wan Animate 2 API anchors identity to the reference character image and transfers poses, expressions, gestures, and full-body motion from the driving video. During complex dance, fast movement, and subtle performance details, it retains facial traits, body proportions, and character style so identity and motion remain coherent throughout the generated video.

  • Anchor visual identity to the reference character image
  • Transfer complex poses, expressions, and full-body motion
  • Maintain character consistency throughout the video

What Can You Create With the Wan Animate 2 API

01

Digital Avatars

Drive a digital presenter from recorded human movement and expressions while preserving the avatar's defining appearance. This creates a direct path from a real performance to reusable digital-human animation.

02

Live-Streaming Hosts

Connect continuous driving footage to the Lite streaming architecture to animate a virtual host with responsive body language and facial performance for live presentation workflows.

03

Interactive Virtual Environments

Translate performer or participant motion into a persistent virtual character for interactive spaces that depend on responsive gestures, expressions, and continuous movement.

04

Film and Animation Production

Use recorded acting as a motion reference for film, animation, and virtual production. Teams can explore a performance on different character designs and camera viewpoints before final production.

Wan Animate 2 vs Wan Animate: Character Animation Comparison

Both models animate a reference character from a driving video. Wan Animate 2 extends the workflow with direct driving-video conditioning, broader cross-character demonstrations, text-controlled viewpoints, and a Lite streaming architecture.

CapabilityWan Animate 2Wan Animate
Task scopeEnd-to-end character animationCharacter animation and replacement
InputsReference image, driving video, text descriptionCharacter image and driving video
Motion workflowDirect driving-video conditioning in a redesigned Diffusion TransformerMotion and expression replication from a driving video
Identity preservationDemonstrated across people, cartoons, robots, and animalsMaintains the supplied character identity during animation
Camera viewpointText-controlled azimuth and elevation with optional Viewpoint LoRAFollows the driving video's viewpoint
Streaming architectureWan Animate 2 Lite reaches 24 FPS at 400×720 on a four-GPU research pipelineStandard offline character animation workflow

The Wan Animate 2 streaming figure refers to Wan Animate 2 Lite in the published four-GPU research setup.

Frequently Asked Questions About Wan Animate 2 API

What is Wan Animate 2?

Wan Animate 2 is an end-to-end character animation framework from Alibaba's Tongyi Lab. It transfers a driving video's body movement, expressions, and fine performance details to a reference character through a redesigned Diffusion Transformer.

How does Wan Animate 2 create character animation?

Wan Animate 2 directly conditions generation on the driving video's latent representation. It preserves frame-aligned performance information without relying on a separate skeleton, keypoint, or intermediate motion extractor.

What inputs does the official Wan Animate 2 pipeline use?

The official pipeline uses a reference character image, a driving video, and a text description of the character and background.

Which character styles can Wan Animate 2 animate?

Wan Animate 2 demonstrates motion transfer for people, cartoon characters, robots, and animals. It adapts the driving performance while preserving the reference character's defining appearance, proportions, and body structure.

How does Wan Animate 2 control the camera viewpoint?

Its optional Viewpoint LoRA uses text descriptions for camera azimuth and elevation. The report demonstrates side, top, bottom, and eye-level viewpoints that differ from the driving video.

What is the Wan Animate 2 Lite real-time result?

Wan Animate 2 Lite reaches 24 frames per second at 400×720 resolution in the published four-GPU streaming pipeline. These values describe the model's documented research configuration.

How is Wan Animate 2 different from Wan Animate?

Wan Animate supports character animation and character replacement from a character image and driving video. Wan Animate 2 adds direct driving-video conditioning, text-controlled camera viewpoints, broader cross-character demonstrations, and a Lite architecture with a published real-time streaming result.

Why Follow the Wan Animate 2 API on PoYo

01

Understand the Core Creative Capabilities

Review motion transfer, identity preservation, camera control, and streaming architecture in one focused overview built around practical character animation decisions.

02

Connect With the Wan Animate Workflow

Compare inputs, task scope, motion handling, and viewpoint control with Wan Animate to see how the character animation workflow develops from one generation to the next.

03

Evaluate It With Other Video Models

Place Wan Animate 2 in the wider PoYo video model catalog and identify the workflow that best matches your character assets, motion references, and production goals.

04

Plan a Clear Path to Production

Use the capability overview, model comparison, and creation scenarios to map how Wan Animate 2 fits your character assets, motion references, and production pipeline.