You Can Make 3D Animation Without After Effects—the Real Problem Is Revising the Second Version

VeytrixAI has posted several tweets over the past two days showing the results of making 3D animations with Claude Opus 5.5. The author puts it directly: one prompt, no After Effects needed.

In the first one, a PDF plus a few instructions produced a finished piece in under an hour. The visuals are not complicated: on a wooden table is an orange cube character, frowning, and next to it is a phone chat interface. Sam asks lunch at 1?, Jo replies sure!, and Mia says saving my cookie for later. The character's expression, the direction of the lighting across the tabletop, and the pacing of the camera push-in are all there.

Another one is even more direct: just one prompt. A pink cube character with heart-shaped eyes reaches for something in the kitchen. The author's emphasis is not on the model itself, but on lighting, motion, and animation quality all holding up at the same time. He asks: how much longer until AI completely transforms motion design?

The third one gives the full pipeline: Claude Opus 5.5 handles generation, Blender handles the scene, and Remotion handles assembling the timeline with code. Characters, lighting, camera movement, and editing are put into the same workflow—no studio rental, no team assembly.

Looking at all three together, there is something beyond the visuals: the workflow has been flattened.

In the past, making ten seconds of 3D animation bottlenecked on people. Modeling, lighting, keyframing, and post-production each handled a stage, and each stage had to be aligned once. Now you put in a PDF or a piece of text, and those alignment meetings in between disappear. VeytrixAI's phrase no expensive studio required is actually referring to this.

There are two common voices in the comments: one side shouts that artists are going to lose their jobs, the other says this is just a toy. Both sides miss the point.

The point is the second version.

Between the impressiveness of a one-shot generation and an asset you can actually deliver, there is a layer of work that usually nobody puts in the video. When the client says move the character a little to the left, make the light a little warmer, cut two seconds from this section—can you make the changes in three minutes, or can you only regenerate? The under an hour in the first tweet is very appealing, but nobody says how long five revisions take. The fact that Blender and Remotion appear in the workflow itself shows the author knows that prompts alone are not enough and that in the end it still has to land on controllable nodes.

The current position is roughly this: generation handles the draft, and tools handle the final tightening. After Effects has not been replaced; it has been moved to the second half.

Every year around this season, people ask what AI will be able to do in a year. Looking at demos makes it easy to overestimate; looking at workflows makes it easy to underestimate. The real signal will have to wait for another kind of video: someone willing to talk about what remains of the same animation after several later revisions.