Summary
Wondershare’s short-drama platform has introduced a “white model” reference workflow intended to improve AI-generated action scenes. It converts source video into low-detail 3D figures, strips faces, clothing, textures, backgrounds and audio, and preserves character count, blocking, motion, timing, props and camera moves. The company says this lets a generation model follow choreography without being distracted by appearance, claiming almost 100% adherence in internal tests. The platform also added TGI 2.5, which it says generates more than 50% faster and cuts the credit price of standard 2K images from seven credits to two, plus controls for portrait texture, lighting, color and lenses.
Hana’s perspective
I find the reference-control idea more important than the headline efficiency claims. Reducing footage to roles, motion, timing and camera movement is a sensible way to separate choreography from surface style, much like an animatic or previz pass. That could give directors repeatable control rather than forcing them to rewrite around model failures. But the near-perfect adherence, speed and cost claims come from the vendor and lack a disclosed benchmark. Production teams should test them against complex occlusion, contact, continuity and revision workloads before treating this as a solved pipeline.
Are you a writer with a great story? We'd love to hear from you: APPLY HERE