Wan2.2 Animate Character

Bring a character to life using real performance video as the motion source.

Upload a still image of your character and a video of someone performing the motion you want. Wan 2.2 Animate extracts pose, facial expressions, and body movement from the video, then applies them to your character image. The output is an animated clip where your character moves, talks, and gestures like the performer in the reference video.

Works with 2D cartoons, 3D renders, illustrated characters, and photographs. Upload your image and video, write a short prompt, hit run.

How do you animate a character with Wan 2.2 Animate?

Upload a character image and a driving performance video. Write a prompt describing the subject. Wan 2.2 Animate extracts pose keypoints and facial landmarks from the video, then generates animation of your character following that motion. An optional refiner pass sharpens detail. Output is 720p at 16 FPS in H.264 MP4.

Load Image Upload a still of the character you want to animate. Front-facing or slight angle works best. The face and body should be clearly visible. Works with photos, illustrations, 3D renders, anime characters, and AI-generated portraits.

The composition doesn't need to match the first frame of your driving video, though aligning them improves results. If your video starts with a medium shot, a medium-shot character image will produce a more natural match.

Load Video Upload the driving performance video. This is the motion source. Wan 2.2 Animate reads body movement, head motion, facial expressions, and hand gestures from it.

Frame load cap defaults to 81 frames at 16 FPS (about 5 seconds). Adjust frame_load_cap to control how many frames the model processes. More frames means a longer output clip but longer generation time.

Clear, single-person footage with visible face and hands transfers cleanest. Cluttered backgrounds or fast overlapping motion make pose extraction harder.

Prompt Describe the subject. "3D man talking" is the default. Simple prompts work well for straightforward animations. For complex scenes, add details about background, action, or style: "Anime girl in school uniform, classroom background, speaking expressively."

The prompt describes the character and scene. The motion comes from the reference video. You don't need to describe the action.

Negative prompt Pre-loaded with quality and artifact filters in Chinese (the language Wan was trained in). Covers overexposed, static, blurred, deformed faces, extra fingers, and similar issues. No need to edit this.

Resolution (default: 832x480) The WanVideoAnimateEmbeds node generates at 832x480 internally. The UltimateSD upscale pass brings the output to 1280x720 (720p). Lower resolutions may fail to capture facial performance details. 720p is the recommended output.

Optional Refiner Enable the Optional Refiner group to process the output with the Wan 2.2 Low-Noise model (T2V-A14B-LOW). This adds a second pass that sharpens detail and reduces noise at the cost of extra generation time. In many cases, the base generation looks good enough without it. Enable it when you need a polished final render.

LoRAs Two LoRAs load by default:

WanAnimate Relight LoRA (strength 1.0): Improves lighting consistency between the character image and the animated output. Keeps skin tones and environmental lighting stable.

Lightx2v distillation LoRA (strength 1.2): Reduces inference steps for faster generation while maintaining quality. This is why the sampler runs at only 6 steps.

Sampler DPM++ SDE sampler, 6 steps, CFG 5. Pre-tuned for the distillation LoRA. The low step count keeps generation fast without sacrificing quality on the Animate model.

What is Wan 2.2 Animate good for?

Wan 2.2 Animate is built for transferring real human performance onto illustrated, 3D, or AI-generated characters. The motion comes from actual footage, so it captures the timing, weight, and expression of a real performance. This is different from text-driven video generation where the model invents the motion.

2D and 3D character animation. Have an illustrated character or 3D render? Film yourself performing the action, then transfer that motion onto the character. Head turns, hand gestures, facial expressions, and body weight all carry through. Faster than keyframing by hand for short sequences.

Talking and presentation clips. Film someone speaking to camera, then transfer the performance onto an AI character, brand mascot, or digital avatar. The face and mouth movement follow the driving video. Good for explainer content, social posts, and character-driven narratives.

Character swapping in video. Have footage of one person but want a different character in the shot? Upload the target character image and the original footage. Wan 2.2 Animate replaces the person while keeping the original performance.

Motion reference testing. Film several takes of a performance, then test each against your character to find the best match. The 6-step distilled sampler generates fast enough to iterate on multiple references in a session.

Honest limitations. Facial performance capture works best at 720p or above. Lower resolutions lose mouth and eye detail. Fast overlapping hand gestures can produce artifacts. Multi-person scenes are not supported in this single-character workflow. Clips beyond 5-8 seconds increase the chance of temporal drift. For longer sequences, generate in segments.

FAQ

What kind of driving video works best with Wan 2.2 Animate?
Clear, well-lit footage of a single person with visible face and hands. Moderate motion transfers cleanest. Film at chest-up or full-body framing depending on what your character image shows. Avoid fast overlapping gestures, heavy occlusion, or multiple people in the frame. 16 FPS is the target frame rate.

Does the character image need to match the driving video's framing?
Not required, but it helps. If your driving video is a medium shot, a medium-shot character image produces a more natural match. The workflow extracts pose keypoints from the video and maps them onto the character, so mismatched framing works but may produce less accurate body proportions.

Should I enable the Optional Refiner?
Enable it when you need a polished final render with sharper detail. Skip it for iteration and testing. The refiner adds a second pass with the Wan 2.2 Low-Noise model, which takes extra time. In many cases, the base output looks good enough for social content and previews.

How long can the output clip be?
Default is 81 frames at 16 FPS (about 5 seconds). Increase frame_load_cap for longer clips. 5-8 second clips have the best stability. Longer clips increase generation time and risk of drift in facial features. For extended sequences, generate in segments and edit them together.

Can I use this for anime or illustrated characters?
Yes. Wan 2.2 Animate works with 2D illustrations, anime characters, 3D renders, and AI-generated portraits. The pose extraction reads motion from the driving video regardless of the character's art style. Illustrated characters with clearly visible facial features produce the best expression transfer.

How do I run Wan 2.2 Animate online?
You can run Wan 2.2 Animate online through Floyo. No installation, no setup. Open the workflow in your browser, upload your image and video, and hit run. Free to try.