Head to head: Bytedance Seedance V1 Pro Fast Image To Video vs Wan 3.0 Prime
Bytedance Seedance V1 Pro Fast Image To Video vs Wan 3.0 Prime
By Ryan Merket · Published · Updated
A 12-clip image-to-video matchup tests whether polished motion or closer prompt execution matters more across rain, racing, a lighthouse, and a fizzy drink. The results hinge on how consistently each model gets the requested scene and action on screen.
Wan 3.0 Prime takes the saved verdict, with aggregate scores of **30.5 to 23.5** and **11 task wins to one**, with no ties. That makes this a clear editorial result on the saved task sample—not a claim about statistical significance or broader model capability.
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 depicts dark, fizzy soda with a convincing splash and a cube rising to float, though an extra cube appears despite the single-cube prompt. Model A has attractive kitchen lighting but looks like clear liquid, and its cube remains at the bottom rather than bobbing with believable buoyancy. (Second judge pass, order swapped — scores are the average of both: Model B shows soda, a splash, and rising fizz, though it appears to contain more than one ice cube. Model A has a polished kitchen sett | Anthropic: Claude Fable 5.1: Model A uses clear water rather than soda and t…
Temporal consistency
A man in a yellow raincoat walking toward camera down a rainy street; his face, coat, and umbrella must stay perfectly consistent with no morphing or flicker from the first frame to the last, steady tracking shot, 16:9.
OpenAI: GPT-6 Luna: Model A shows a clear full-body walk toward the camera while keeping the man, coat, and umbrella visually consistent. Model B has more cinematic rain and lighting, but its close framing makes the walking motion much less apparent, despite strong consistency across frames. (Second judge pass, order swapped — scores are the average of both: Model B has a polished, clearly rainy city setting and a stable face, coat, and umbrella, but the man appears nearly stationary across the frames. Model A show | Anthropic: Claude Fable 5.1: Model A shows genuine walking motion with a sta…
Salt flat pursuit
One continuous 7-second shot at sunrise over the Salar de Arizaro salt flat: a low helicopter-style camera races just off the left rear quarter of a cobalt-blue open-wheel land-speed car numbered 74 as it blasts forward, the lens subtly pushes closer while matching speed, the background mountains smear with velocity and heat shimmer, rooster tails of white salt dust stream behind the tires, suspension jitters over tiny ripples, and brief realistic motion blur streaks the spinning wheels and ground while the car stays crisply dominant in frame; amber sun skims the reflective crust, long shadows knife sideways, and the mood is ferocious, exultant momentum, 16:9
OpenAI: GPT-6 Luna: Model B better matches the requested left-rear-quarter pursuit, with mountains, reflective salt, and strong ground streaking conveying speed. Model A keeps the car larger and crisper and has a subtle push-in, but its front-side view and comparatively subdued landscape motion miss more of the specified shot. (Second judge pass, order swapped — scores are the average of both: Model B better matches the requested rear-quarter tracking view, open-wheel blue car, salt-flat mountains, and sunrise refl | Anthropic: Claude Fable 5.1: Model B nails the prescribed left-rear-quarter…
Cliff kite occlusion
A single continuous 6-second shot over a windswept basalt headland at blue hour: the camera glides sideways in a smooth drone-like track along a sea cliff while a battered saffron research kite shaped like a manta ray, with a missing notch in its left wingtip and a string of three tiny silver sensor pods under its belly, dives across frame and passes completely behind a tall white rotating lighthouse lantern room for a full beat, then re-emerges on the far side with the same torn wingtip, same pod spacing, and same taut line angle, continuing its arc above foamy black water; cold beacon flashes rake the mist, the horizon glows dim violet, and the mood is eerie but precise, 16:9
OpenAI: GPT-6 Luna: Model A more clearly shows the kite passing behind the lighthouse and re-emerging, which is the prompt’s defining action, though its lighting is brighter than the requested blue hour. Model B has a more convincing eerie blue-hour atmosphere, but the kite passes in front of rather than fully behind the lighthouse. (Second judge pass, order swapped — scores are the average of both: Model A more clearly depicts the kite passing behind the lighthouse lantern room and re-emerging, the prompt’s key ac | Anthropic: Claude Fable 5.1: Model A attempts the lantern-room occlusion but…
Matchup powered by OpenRouter.