Wan 2.1 Mocha

Frequently Asked Questions

What is Wan 2.1 Mocha?

MoCha is an end-to-end video character replacement system designed to swap a character in a source video with a new character (provided via reference images), without relying on explicit structural guidance (e.g., pose/depth maps) for every frame.

How are credits calculated?

Duration- and resolution-based: 480p ≈ 12 credits/s, 720p ≈ 24 credits/s. Minimum charge is 60 credits (5 seconds at 480p). Maximum video length is 120 seconds.

What are the use cases for character replacement?

Creating personalized content, character animation for games/films, translation/localization with different actors, creating demo content, replacing actors in existing footage, and generating character variations for storytelling.

What are the key features of MoCha?

End-to-end character replacement without pose maps, natural expression transfer, holistic movement replication, maintains character appearance consistency, supports various resolutions (480p/720p), and preserves motion dynamics from source video.

How do I get best results?

Match composition & pose between reference image and video, keep aspect ratios consistent, use clear reference photos (PNG/JPG), ensure camera position and human pose align, and use optional prompts to preserve specific elements like outfit or expression.

What inputs are required?

Character image (PNG/JPG recommended) and motion video are required. Optional: prompt for guidance rules, resolution selection (480p/720p), and seed for reproducibility.

Can I control which character gets replaced?

MoCha automatically identifies and replaces characters based on the reference image. For better targeting, ensure your reference image and video share similar composition, camera angle, and pose.

What video formats are supported?

Input: MP4 or WebM. Output: MP4 format. Maximum input video length is 120 seconds.

How does MoCha differ from other character animation tools?

MoCha is specialized for character replacement in existing videos, not generating new motion. It transfers a character's appearance to another character's motion without requiring pose estimation or depth maps, making it efficient for production workflows.