Descriptions:
Youri van Hofwegen demonstrates a workflow for automating AI movie production inside Imagine Art’s node-based canvas — a visual interface resembling tools like n8n or ComfyUI — that lets creators build an entire multi-scene film pipeline and run it in one pass. The core problem addressed is speed: standard AI filmmaking requires generating each clip individually, which is time-consuming and breaks creative flow.
The tutorial walks through two production setups. The first chains a prompt node, an image node (using Nano Banana Pro at 4K for maximum detail), and a video node (V3 for human subjects, or Kling 2.1 Pro) to generate a single cinematic scene of a Starship engineer in a maintenance corridor. The second setup scales this into a full five-scene movie by passing the end frame of each clip as an image URL reference to the starting frame input of the next scene, creating visual consistency across cuts without manual intervention.
Key model selection guidance is included: Imagine Art 1.5 and Nano Banana Pro for stills, V3 when scenes feature people (Sora 2 is excluded for human subjects), and Kling 2.1 Pro for motion quality. Images are generated at 2K rather than 4K when Kling is the downstream video model, as it does not support 4K inputs. Once the workflow structure is built once, creators can swap in new prompts for entirely different stories without rebuilding the pipeline.
📺 Source: Youri van Hofwegen · Published December 10, 2025
🏷️ Format: Hands On Build
