Same Prompt, Fable Gives a Big Bird on a Bike a Sense of Physics

Chetaslua posted a four-panel video: four clothed big birds riding bicycles. The core message in the caption is direct: with the same prompt, Fable delivered a hyper-realistic result with rich detail and an artist-like sense of physics, without any extra requests.

In the video, the big bird in the top left wears a red scarf, with a sunset seascape and lighthouse in the background; the top right one wears a green hat and red scarf, with a sunrise beach and sun behind it; the bottom left one wears orange gloves, riding along the coastline under a clear sky with white clouds; the bottom right has no background, just a gray big bird riding a simple bicycle. Some commenters call out the feathers and bicycle chain, saying the textures are fierce.

The tweet pushes the conclusion further: video or image models will no longer be needed. In the replies, conclusions arrive faster than frames render.

Some ask whether this is shadow routing, implying it is routed to another model behind the scenes; others ask whether it is still improving this week or is the same checkpoint. Some care about access and pricing. Others suspect Fable may have received a reference image, or that Fable 5.5 did not follow instructions and simply switched the subject and style.

The more specific doubts focus on comparability. Some say same-prompt comparisons tend to reward over-delivery: hyper-realism is great when nobody asked for it, but not necessarily when you really want cartoons. Some want to see whether it holds up under a style brief. Others point out that same-prompt tests mostly show a model's default style; the real gap is keeping the same character across ten frames. Text rendering and precise local editing are still where image models often fail, and specialized models may continue to have the edge. Hands and fast motion have not been shown yet.

On the other side, some argue that if the same model does both images and video, half of AI video startups will become features. That is not a light judgment.

This video shows that Fable is strong on default aesthetics, materials, and physical priors. To replace video models, it still lacks publicly reproducible conditions: who can use it, how much it costs, whether the checkpoint is stable, whether style instructions hold, and whether characters and local edits are controllable. The conclusion can be exciting, but do not rush to send video models away just yet.