Head to head: H3 Max Hand Drawn vs Vidu Start End to Video
H3 Max Hand Drawn vs Vidu Start End to Video
By Ryan Merket · Published · Updated
H3 Max Hand Drawn and Vidu Start End to Video take different approaches to short-form motion, from a scarfed whippet crossing a rooftop laundry line to crowds and a bursting water balloon. This head-to-head weighs both visual style and how clearly each clip carries out the requested action.
The saved result is a tie on this task sample, despite Vidu’s higher aggregate score: **24.6** for H3 Max Hand Drawn and **28.2** for Vidu Start End to Video. Across the 12 task outcomes, H3 won 4, Vidu won 7, and 1 was tied—but those wins divide along distinct prompt types rather than making a decisive case for either model overall. H3 had the clearest edge on the whippet prompts: it more consistently showed the dog passing behind the sheets and re-emerging with its scarf and ring. Vidu’s rooftop clips often looked more polished, but the dog sometimes stayed visible beneath the laundry or moved in the wrong direction. Crowd motion and fluid dynamics leaned Vidu: its Tokyo scenes felt more specific and its water bursts more visually striking, though dense pedestrians could blur together and some burst sequences were inconsistent. The goat-and-satchel prompt split, with each model better conveying the action in some trials and neither consistently nailing it. There were 3 successful repeated trials. The mean paired score difference (A−B) was -3.633333333333333, with a Student-t 95% confidence interval from -7.442119472230537 to 0.175452805563872. That interval describes this sample; it is not a probability of either model winning or of being generally better. **Final call: Too close to call (not statistical). No overall winner—this assessment applies only to the saved task sample.**
Whippet behind laundry line
In a breezy rooftop garden at blue hour, a sleek silver whippet wearing a mustard knit scarf trots toward camera carrying a neon green rubber ring, then passes fully behind a row of flapping striped bedsheets on a laundry line and re-emerges on the other side still wearing the same scarf and holding the same ring without distortion or substitution, continuing its trot as if uninterrupted; the camera tracks left alongside the dog in a smooth lateral move with slight handheld bob, maintaining the single continuous shot as planters, string lights, and damp concrete catch the cool evening glow, creating a quiet, slightly suspenseful mood, 16:9
OpenAI: GPT-6 Luna: Model A better matches the blue-hour rooftop setting and shows the whippet approaching with the same scarf and ring, though it does not clearly complete the requested pass behind the sheets and re-emergence. Model B is visually polished and shows movement behind laundry, but the dog travels away from the camera and the scene feels more like night than blue hour. (Second judge pass, order swapped — scores are the average of both: Model A more clearly shows the whippet approaching camera with the | Anthropic: Claude Fable 5.1: Model A follows the prompt closely—blue-hour roo…
Crowd motion
A busy Tokyo scramble crossing seen from above, dozens of pedestrians crossing in different directions, each moving independently without merging or warping into one another, overcast daylight, 16:9.
OpenAI: GPT-6 Luna: Model A is stylized and less photorealistic, but its pedestrians remain individually legible and move without obvious merging. Model B more clearly depicts a real Tokyo scramble crossing and a much larger crowd, but the dense motion becomes a blurred, clumped mass rather than independent pedestrians. (Second judge pass, order swapped — scores are the average of both: Model B closely matches the elevated, overcast Tokyo scramble crossing, with dense crowds moving across several crosswalks; some g | Anthropic: Claude Fable 5.1: Model A keeps dozens of clearly separated pedes…
Fluid & particle dynamics
Cinematic slow-motion of a water balloon bursting, the sheet of water expanding and droplets scattering through the air in convincing detail against a dark background, hard side light, 16:9.
OpenAI: GPT-6 Luna: Model A clearly shows a balloon before it bursts, but the splash is muddy and the pale floor weakens the dark-background look. Model B has a more convincing, beautifully lit expanding sheet and scattering droplets, though it omits the balloon itself. (Second judge pass, order swapped — scores are the average of both: Model B delivers a striking, detailed expanding water sheet and scattered droplets against a dark background, with coherent progression across the frames, though the balloon itself | Anthropic: Claude Fable 5.1: Model A renders the burst as a hand-drawn, penci…
Goat fastens courier satchel
A scruffy cinnamon-and-white miniature goat named Pesto stands on a sun-warmed houseboat deck and, with surprising dexterity, hooks a brass snap buckle closed on a tiny canvas courier satchel hanging from its chest, then nudges the flap flat with its muzzle while shifting its hooves for balance; the camera starts low at hoof level and makes a slow handheld push-in with a slight arc to the right, catching the full body movement in one continuous shot; late-afternoon amber light glints off chipped teal railings, the river behind softly sparkling, and the mood is playful but observational, like an energetic documentary, 16:9
OpenAI: GPT-6 Luna: Model A better captures the scruffy goat, sparkling river, amber light, and apparent slow push-in as the goat noses the satchel flap. Model B looks more photorealistic, but the goat is cleaner and the sampled frames show little of the requested fastening action or camera movement. (Second judge pass, order swapped — scores are the average of both: Model B has a convincing sunlit houseboat setting and full-body goat, but shows little of the requested fastening or muzzle action. Model A is stylize | Anthropic: Claude Fable 5.1: Model A nails the scruffy cinnamon-and-white go…
Matchup powered by OpenRouter.