AI Video Is Growing Up — From Generating Clips to Directing Productions

iQIYI and Alibaba’s Wan3.0 point to the next phase of AI video: moving beyond impressive clips generated from prompts toward production environments where creators can plan scenes, control cameras, maintain greater visual consistency and direct AI across a connected workflow. For marketers and agencies across Asia-Pacific, that could change how video gets made—and where the real creative advantage lies.

Anyone who has experimented seriously with AI video knows the frustration. One clip looks great, the next changes the character’s face, another gets the camera angle wrong, and soon you are generating version after version trying to make everything fit together. That may be fine for experimentation. It is much harder to build professional production around it.

Now, that is beginning to change. Chinese streaming company iQIYI is bringing Alibaba’s Wan3.0 model into NadouPro, its AI platform for professional film and television production, with the capabilities being showcased through a new Global AI Creator Challenge. The competition is the immediate news.

The bigger story is what sits behind it. AI video is beginning to move from the prompt box to the director’s desk.

From prompting to directing

The first wave of generative video was largely about prompting. You described what you wanted and hoped the model interpreted it correctly. If it didn’t, you changed the prompt and tried again.

Professional filmmaking needs more control. Directors have to think about characters, locations, camera positions, movement, timing, continuity, composition and the relationship between one shot and the next. The emerging generation of AI production platforms is beginning to address that gap.

NadouPro, for example, is designed around a broader production workflow that iQIYI says can take creators from scripts and visual assets through storyboarding to the finished film.

Its 3D Directing Studio lets creators stage 3D models and cameras, refine movement and...