Tiangong Short-Drama Workbench Launches Dual-Track Mode: Agent Storyboarding and Infinite Canvas

Core Highlights

Kunlun Wanwei, one of China's prominent internet and artificial-intelligence companies, launched a short-drama workbench under its Tiangong brand and introduced a dual-track creation mode described as agent-powered storyboarding combined with an infinite canvas. The tool targets the two most persistent pain points in AI-generated short dramas today: characters whose faces change mid-scene and camera positions that drift between shots. By adopting a controllable production approach, Tiangong aims to make the mass production of short dramas stable and reliable rather than a lottery of random outputs that must be discarded after rendering. As a significant player in China's domestic AI-video race, Tiangong is breaking the film and television production pipeline into callable modules, which clearly signals the intention of large-model companies to move upstream into content production rather than staying at the level of raw, uncontrolled generation. The move also reflects a wider trend in China, where short dramas have become a massive, fast-moving content category and where producers are hungry for tools that cut production cost without sacrificing watchability. Short dramas, with their fast release cadence and thin margins, leave little room for expensive reshoots, so a tool that prevents inconsistency at the source is directly valuable to producers who must ship episodes on a tight schedule.

What Happened and What It Can Do

The core of the dual-track mode is a director agent that automatically parses the script, plans the positions of characters and the angles of cameras, and supports the generation of multi-view detail images. In the past, AI-generated short dramas frequently suffered from inconsistencies where the same character's face looked different from shot to shot, and where positions jumped unpredictably between scenes. This new mechanism suppresses that drift by locking the storyboard and viewpoints in advance, treating the process as planned shooting instead of free-form generation that changes with every render. Creators no longer need to regenerate an entire scene when one detail is off; they can adjust the plan and re-render only what changed, which is closer to how a real shoot operates. The workbench ships with cinematic prompt templates, 720-degree panoramic images, and a 3D director console, allowing creators to arrange scenes within a unified space and keep visual continuity under control. In simple terms, it transforms 'random generation' into 'shooting according to a plan', which is what professional crews actually need to deliver a finished work.

Technical Details

The tool uses the director agent as its central hub. It first reads the script and then produces a storyboard plan, after which it generates multi-view detail images to guarantee character consistency across the whole clip. The 720-degree panoramic images and the 3D director console provide spatial references so that camera positions and character placements can be planned precisely instead of being left to chance. Keeping a single source of spatial truth means the agent can reason about the scene the way a human director would, rather than treating each frame as an isolated image. By packaging prompt engineering, spatial modeling, and agent orchestration into ready-made templates, the product lowers the trial-and-error cost for non-professional creators and brings the output closer to the controllable standards of real film production, where consistency is not optional but a basic requirement.

Comparison with Competitors

Compared with overseas video models that are known for impressive single-shot generation, Tiangong emphasizes the completeness and controllability of the production pipeline. It does not chase a single stunning generation; instead it sells stability and consistency as its main advantages, which matter far more when a drama has dozens or hundreds of shots that must stay coherent. Where many tools optimize for the wow factor of a single clip, Tiangong optimizes for the boring reliability that a production schedule actually requires. Amid the domestic Chinese short-drama export boom, this kind of tool built for industrial-scale production exactly fills the engineering gap that creators face between an initial idea and a finished clip that can actually be released to an audience.

Industry Impact and Use Cases

Three works have already been published on DramaWave and reached million-dollar-level revenue within seven days, which validates the commercial viability of this path with hard market results rather than promises. The seven-day revenue result also lowers the risk for investors, who can now back AI-produced content with more confidence that a series will stay visually consistent from the first episode to the last. For small and mid-sized teams, controllable AI short-drama production means lower cost and faster iteration, letting them test many concepts without hiring large crews. For platforms, a stable supply of consistent content supports scaling and predictable scheduling. If the DramaWave results hold, they suggest that controllable AI production is not just a technical curiosity but a revenue engine, which will pull more studios toward agent-assisted pipelines. As AI video moves toward industrialization, competition at the tool layer will gradually shift from 'can it generate' to 'can it produce stable, controllable output at scale', and Tiangong is positioning itself early on that curve.