Alibaba's Qwen Creation launches Agent Teams for AI short-drama and video production
The AMW Read
Incremental feature update for an already-tracked player, packaging multi-agent orchestration with existing Wan video models rather than introducing a new capability or player.
Alibaba's Qwen Creation launches Agent Teams for AI short-drama and video production
Alibaba's Tongyi Qianwen content platform, Qwen Creation (千问创作), has rolled out an "Agent Teams" feature that coordinates multiple specialized AI agents — acting as writer, director, and art director — to carry a video project from concept through finished cut. The feature ships alongside first access to the Wan 3.0 and Wan 3.0 Prime video-generation models plus the Qwen-Image-3.0 model, and is aimed at AI short dramas, comic-style dramas, marketing ads, and image design. Users submit a single creative brief; the system plans the workflow, assigns agent roles, and lets users adjust direction mid-production. The feature is live in closed beta on Qwen Creation's PC site and the Qwen app, with multi-device support, and is not yet open to all users.
The launch places Alibaba's video-generation stack and image model behind an agentic production layer rather than shipping them as standalone generation tools, reflecting a broader shift toward multi-agent orchestration in creative production, where discrete specialist roles — not a single prompt-to-output call — assemble a finished asset. It also sharpens Alibaba's position in China's fast-growing AI short-drama tooling market, where lowering the cost and skill floor for scripted video content is the main competitive axis against other domestic pipelines.
For builders, the release signals that packaging generation models into role-based agent pipelines, rather than exposing raw text-to-video APIs, is becoming the differentiator in vertical content tooling. For investors tracking China's AI content stack, Qwen Creation's beta-stage rollout is worth watching for conversion into paid or ad-supported short-drama production once it exits closed testing.
