Head to head: Bytedance Seedance V1 Pro Fast Image To Video vs H3 Max VHS
Bytedance Seedance V1 Pro Fast Image To Video vs H3 Max VHS
By Ryan Merket · Published · Updated
Four image-to-video challenges put the models through rainy neon street action, a liquid-and-ice physics test, a ramen camera orbit, and a bakery courier follow. The comparison focuses on prompt fidelity, motion, and visual consistency across three repeated trials.
H3 Max VHS takes this matchup decisively on the saved task sample: it won all 12 task outcomes, with no ties, against Bytedance Seedance V1 Pro Fast Image To Video. The aggregate scores were **29.5** for H3 Max VHS and **18.4** for Seedance—a clear result in this sample, not a statistical claim about broader capability. The gap came from getting the requested action on screen. H3 more often showed the lime-green bicycle moving through a puddle, a single ice cube rising in dark fizzy soda, a visible orbit around steaming ramen, and a courier running with stacked pink boxes. Seedance often looked polished and sharp, but its scooter stayed nearly parked, its drink read as clear water with extra cubes, its camera barely moved, and its courier walked with boxes side by side. H3 was not flawless: some splashes and liquid motion were understated or exaggerated, the orbit did not always reach the requested sweep, and details varied. But across these prompts, it repeatedly delivered the defining action while Seedance’s cleaner presentation could not make up for missing it. Across three successful repeated trials, the reported mean paired score difference (A−B) was -11.083333333333334; its Student-t 95% confidence interval was -12.82813432036058 to -9.338532346306089. That interval describes the mean paired difference, not a probability of winning or a claim about general superiority. **Final call: H3 Max VHS is the clear winner on this saved task sample, chiefly because it follows the requested motion and scene action more faithfully.**
Rain-gutter neon burst
A single continuous 16:9 shot on a drizzly side street behind a discount electronics shop at blue hour: the camera glides low and slightly sideways along the curb, tracking a dented lime-green delivery bike as it cuts through a deep rain gutter puddle, while a sheet of water sprays outward, ripples race along oil-slick asphalt, droplets bead and fling from the spinning spokes, and steam curls from a nearby manhole, all moving with convincing fluid dynamics under flickering magenta pharmacy signage, with a tense, electric city-night mood.
OpenAI: GPT-6 Luna: Model B shows a lime-green bicycle moving through the gutter with a visible splash, steam, wet reflections, and a low tracking perspective that closely match the prompt. Model A is polished and colorful, but its stationary scooter and minimal water action miss the central motion of the scene. (Second judge pass, order swapped — scores are the average of both: Model B closely matches the low, rainy curbside scene, with a lime-green bicycle moving through a puddle, visible spray, neon reflections, | Anthropic: Claude Fable 5.1: Model A shows a parked mint scooter (not a dent…
Physics realism
A close-up of a single ice cube dropped into a tall glass of soda: it falls, splashes, bobs, and settles with believable buoyancy and fizzing bubbles rising naturally, soft kitchen light, 16:9.
OpenAI: GPT-6 Luna: Model B better depicts dark soda with abundant rising fizz and a single ice cube buoying toward the surface, though the drop and splash are not shown clearly. Model A is brighter and more polished, but the cube appears to duplicate and the liquid lacks convincing soda fizz or splash motion. (Second judge pass, order swapped — scores are the average of both: Model B shows dark soda, a single ice cube rising toward the surface, and visible fizz, though the splash is understated. Model A is polishe | Anthropic: Claude Fable 5.1: Model A renders clear water being poured rather…
Camera motion control
A single smooth orbit (roughly 180 degrees) around a steaming bowl of ramen on a wooden table, the camera circling at a constant speed and height in one continuous move, warm restaurant light, 16:9.
OpenAI: GPT-6 Luna: Model B shows clearer, continuous-looking orbiting motion, with the bowl’s toppings and background shifting in perspective while the warm restaurant atmosphere remains consistent. Model A is bright and appealing, but its viewpoint changes less noticeably and reads closer to a mostly static shot. (Second judge pass, order swapped — scores are the average of both: Model B shows a coherent changing viewpoint around the steaming ramen while maintaining the warm restaurant setting and a consistent su | Anthropic: Claude Fable 5.1: Model B shows a more convincing orbital move: t…
Bakery runner follow
A single continuous 16:9 shot in early golden morning on a narrow avenue lined with produce stands: the camera jogs backward at chest height to keep a young bakery courier in a mustard windbreaker centered and sharp as she hustles downhill carrying two stacked pink pastry boxes, weaving around a newspaper rack and a silver fire hydrant while buses slide past, awnings sway, pedestrians cross behind her, and reflections skim across shop windows, with warm side light and an urgent but buoyant everyday-city mood.
OpenAI: GPT-6 Luna: Model B better captures the courier running with stacked pink boxes in warm morning light, with buses, produce stands, and a hydrant in the scene. Model A is sharper and consistent, but the courier appears to walk and carries the boxes side by side rather than stacked. (Second judge pass, order swapped — scores are the average of both: Model B more clearly shows the courier running with the pink boxes stacked, backed by warm light, produce stands, buses, and a hydrant. Model A is sharper and vis | Anthropic: Claude Fable 5.1: Model B delivers the urgent running courier wit…
Matchup powered by OpenRouter.