Kiwikiwi Officially Launches: Bringing Control to AI Filmmaking — One Set of Assets, Two Creative Possibilities: Films and Interactive Games
张雨晨(Aurora Zhang)
张雨晨(Aurora Zhang)

2026-11-02

Kiwikiwi Officially Launches: Bringing Control to AI Filmmaking — One Set of Assets, Two Creative Possibilities: Films and Interactive Games

As AI short-drama production surges to tens of thousands of titles per month, yet the hit rate falls below 0.05%, the industry's dividing line is no longer whether content can be generated, but whether it can be delivered consistently and whether creative assets can be reused.

On October 31, KiwiKiwi, an AI-powered film and video creation platform developed by Xingke Pictures, officially launched, offering two dedicated creative portals for B2B and B2C users: kiwi.film, the B2B enterprise platform designed for professional teams and enterprise-level film production needs; and kiwikiwi.art, the B2C platform that provides individual creators with AI-powered filmmaking services.

The Other Side of the Production Boom: The AI Look and Low Returns

AI filmmaking data in the first half of 2026 reveals a sharp divide. Douyin saw 221,900 new AI-generated animated dramas, while 430,000 micro-dramas were released across the internet, with AI-generated dramas accounting for more than 90%. However, according to DataEye, only 0.47% surpassed 100 million views. Based on the 50-million-view break-even threshold, only around 1.3% reached the target. More than 90% of small and mid-sized production teams failed to generate a profit.

Behind this production boom, the "AI look" has become one of the most common audience complaints. Characters change appearance between shots, props become misaligned, camera movements lack motivation, and disconnected clips create inconsistent pacing. The underlying problem is not simply insufficient image quality, but the lack of physical and identity continuity between shots. A 90-minute feature film generated in 15-second segments would require 359 transition points, each representing a potential break in visual consistency.

Three Types of Discontinuity Three Layers of Solutions

Kiwikiwi does not rely on more powerful models to solve these problems. Instead, it uses engineering workflows to reduce uncertainty before generation.

Identity discontinuity — Characters drift in appearance across different shots. The platform turns characters, scenes, and props into project-level assets. Rather than duplicating images, each shot references the original assets it uses. When an original asset is modified, all shots referencing it are updated accordingly. Character consistency across shots becomes a natural result of the platform's asset management structure.

Spatial and camera movement discontinuity — Props become misplaced and camera movements lack clear intent. KiwiKiwi features a built-in 3D Director's Console, allowing users to freely configure camera positions, focal lengths (12 options ranging from 14mm to 200mm), and six shot sizes. Camera paths support straight lines, smooth curves, and orbit movements, along with four speed curves. Each movable object can have its own keyframe trajectory, with automatic orientation along the path. Once the pre-visualization is approved, users can export a standardized shot list. It serves both as a shooting plan for production teams and as precise input for generative models. This turns camera movement from prompt-based guesswork into exportable parameters.

Temporal discontinuity — Disconnected clips lead to abrupt changes in pacing. The platform provides video extension capabilities, separating the model's maximum generation duration from the intended shot length. Its editing workspace includes a timeline, audio tracks, and transitions, allowing finished videos to be assembled around a deliberate pacing structure rather than simply stitched together.

One Set of Assets Two Revenue Streams

What sets Kiwikiwi apart from a conventional AI tool is its core capability: the same set of assets can be used to create two different products.

Within the platform's data structure, a "shot" is not merely a video file, but a structured production record containing eight fields: its scene, characters, visual description, dialogue, camera position and movement, duration, generation instructions, and referenced assets. The final field is critical to the entire system. Changing an ending, for example, only requires modifying one node and generating a few additional shots. All other shots can be reused directly from the asset library, retaining their original camera settings, characters, scenes, and dialogue.

Based on an 18-shot interactive game scenario, if 12 shots from the original film version are already available, only six additional shots need to be generated for the interactive version. This translates to an asset reuse rate of approximately 67%, with the marginal cost of secondary monetization reduced to just one-third of the original production cost.

Industry data puts the Chinese interactive film and game market at approximately RMB 18–20 billion, yet overall development penetration remains below 10%, while around 90% of interactive game projects reportedly operate at a loss. The biggest obstacle is production cost. Kiwikiwi's asset reuse mechanism directly targets the underlying cost structure.

Visual Quality and Ecosystem: Building an Orchestration Layer, Not Another Model

At the infrastructure level, Kiwikiwi does not develop its own generative models. Instead, it focuses on orchestrating different models within a unified workflow. Video generation integrates engines such as Seedance 2.0, Jimeng 3.0/3.0 Pro, and Agnes Video. Image and 3D generation connect with Seedream and Doubao Seed3D, while world models can directly import Marble from World Labs, which generates navigable, high-fidelity 3D environments from text, images, or videos.

Enterprise customers can also use BYOK (Bring Your Own Key), allowing them to connect their own model API keys and keep usage within their existing quotas. As underlying models evolve weekly, the platform can upgrade its integrations without having to rebuild the entire production layer.

According to Xingke Pictures, Kiwikiwi's goal is not simply to make AI generate better content, but to move AI filmmaking from one-time production to iterative, long-term asset management.

When characters, scenes, and storyboards become searchable and reusable assets, making a film and developing an interactive game are no longer two separate projects, but two different ways of extending the same creative resources.