Descriptions:
Creator Dan Kieft walks through his complete end-to-end workflow for producing a multi-scene AI anime short using Seedance 2.0, Nano Banana 2 (accessed via Hickfield), Claude, Suno for background music, and standard video editing tools. The tutorial tackles the hardest problem in AI video creation — maintaining character consistency across different scenes, environments, and shot types — and presents a reproducible four-step system for solving it.
The process begins with black-and-white manga-page storyboards generated in Nano Banana 2 using detailed tagged prompts that label each character by name within every panel, giving the model precise compositional direction across multiple shots in a single generation. These black-and-white frames are then colorized using the original as a reference, locking in consistent color palettes across the full production. An optional animation notes step adds directorial guidance for motion and camera behavior before the final stage: using the colored reference images to drive Seedance 2.0 video generation. Character consistency is further enforced through Seedance’s “elements” system — a LoRA-style feature where users upload multiple reference images of a character to train a reusable embedding that persists across generations.
Kieft emphasizes the importance of creating separate character sheets for every variant of a character (normal and powered-up states are both covered for the video’s two protagonists, Kai and Saitama), and shares all prompts through a free community resource. The finished anime short, built entirely without manual illustration skills, demonstrates how combining Claude for prompt engineering with Nano Banana 2 and Seedance 2.0 can yield coherent, emotionally resonant AI animation.
📺 Source: Dan Kieft · Published April 29, 2026
🏷️ Format: Tutorial Demo
