The Umbrella Gentleman: Stop-Motion Style Animation Test | Challenges in Character Design

Not every image that looks perfect as a still frame turns into a convincing animation. This short test takes a whimsical skeleton character, dressed in a patchwork umbrella and detailed costuming, and puts it through Runway's image-to-video tool to see what survives the jump from a static picture to motion.

The results are mixed in an instructive way. The background and objects around the character animate smoothly, giving the scene a believable sense of movement and depth. The character's face is a different story. Expressions as simple as a smile or a blink do not translate cleanly, exposing one of the more stubborn limits of current image-to-video generation: faces, especially stylized or non-human ones, are harder to animate convincingly than backgrounds, fabric, or props.

What an animation test is actually for

An animation test is not meant to be a finished piece. It is a short, exploratory sequence used to investigate how a scene, character, or concept behaves once motion is introduced. Think of it as a visual sandbox where the goal is not polish but information: does this style of image respond well to this particular AI tool, and if not, where exactly does it break down. That distinction matters a lot when working with AI-driven text-to-video and image-to-video tools, since different model architectures interpret the same visual prompt in different ways, and a technique that works beautifully on one character design can fail on another.

Where the umbrella gentleman falls apart

The specific failure here is localized and useful to name precisely. It is not the whole animation that fails, it is the facial rigging. Articulated limbs, the patchwork umbrella, and the surrounding environment all hold up under motion, which suggests the model has an easier time animating rigid or semi-rigid elements and textures than it does subtle facial musculature on a stylized skeleton face. That is a helpful data point for anyone designing characters intended for AI animation: faces built for stillness do not automatically translate into faces built for expression.

The real lesson: adapt, don't abandon

Rather than treating this as a dead end, the test frames the outcome as a normal part of the process. If a piece of art is loved for how it looks, that is still reason enough to try animating it. Some images work on the first attempt, others need adjustment, and that is simply how experimentation with these tools goes. The suggested next steps are practical: tweak the animation style, change the prompt, or in some cases redesign the character itself so its features are better suited to the tool's strengths.

Key takeaways

• A visually striking still image does not guarantee it will animate well, especially around facial expressions.
• Backgrounds, props, and articulated costume elements can animate convincingly even when a character's face does not.
• Animation tests exist specifically to surface these mismatches early, before committing to a full production.
• When a face fails to animate, the fix is usually a new prompt, a style change, or a character redesign rather than abandoning the piece.
• Adaptation, not perfection on the first try, is the normal workflow when using AI-driven animation tools.

Try it yourself

If you are experimenting with image-to-video tools like Runway, treat your first pass as a diagnostic rather than a final product. Run a short test on your character, watch closely for where the motion breaks down, and be ready to adjust your prompt or redesign the trouble spots. That habit of testing before committing is the same iterative approach Humanitarians AI uses across its own AI-generated creative work.