Head to head: Kling 4.0 Image to Video vs Wan v2.2 A14B
Kling 4.0 Image to Video vs Wan v2.2 A14B
By Ryan Merket · Published · Updated
A focused comparison of two image-to-video models across underwater camera moves, a small-scale physics scene, character continuity, and controlled camera motion. The verdict weighs prompt execution and visual coherence across repeated trials.
Kling’s aggregate score is **29.7**; Wan’s is **23.3**. But the saved result is a **tie**, with Kling taking 9 task wins, Wan 2, and 1 tie. The editorial strength is **Too close to call (not statistical)**—the higher aggregate does not overturn the saved verdict. Kling’s clearest recurring advantage is camera direction: it more often delivered the requested centered dolly toward the Nautilus dome and a readable orbit around the ramen bowl. In the ice-cube scene, it generally kept to the single-cube prompt and produced more coherent motion, even when its splash was less convincing than Wan’s. Wan had its best showing on the pearl-diver task, winning twice. Its images could be vivid and realistic, but tighter crops and departures from the requested suit, goggles, or framing weakened consistency across those comparisons. Neither model was uniformly reliable across the set. Across three successful repeats, the mean paired score difference (A−B) was 6.366666666666666. The Student-t 95% confidence interval for that mean difference was -3.089657435598858 to 15.82299076893219; this interval is not a probability of winning. This assessment applies only to the saved task sample. **Final call: tie—too close to call.**
Nautilus Dolly-In
A 7-second underwater clip inside a midnight-blue kelp corridor where a brass-riveted research submersible named Nereid-7 glides slowly forward while a cloud of tiny silver baitfish parts around it; execute a smooth slow dolly-in straight toward the pilot dome, closing the distance evenly with no lateral drift or shake, as the sub’s amber instrument lights glow through the glass and narrow shafts of sun ripple down from the surface, creating a tense, exploratory mood, 16:9
OpenAI: GPT-6 Luna: Model A delivers a smooth, centered dolly-in toward the pilot dome, with the baitfish parting around the submersible, amber lights, and kelp corridor all clearly visible. Model B is atmospheric and coherent, but its side-on framing and lateral-looking movement miss the requested direct approach toward the dome. (Second judge pass, order swapped — scores are the average of both: Model A delivers a centered, steady dolly-in directly toward the pilot dome, with the NEREID-7 marking, glowing instrum | Anthropic: Claude Fable 5.1: Model A executes the requested straight dolly-i…
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 A better matches the tall-glass close-up and shows a single cube rising to float with plausible fizz, though the actual drop and splash are not clearly depicted. Model B has a more obvious splash, but the short glass and enormous foamy overflow make the physics and prompt adherence less believable. (Second judge pass, order swapped — scores are the average of both: Model A clearly shows the single ice cube sinking and then bobbing back to the surface, with restrained bubbles and a believab | Anthropic: Claude Fable 5.1: Model A shows a tall glass with the cube plungi…
Consistent Pearl Diver
A 9-second single underwater shot of one pearl diver named Suri Han swimming steadily left to right above a lavender sea-grass meadow, her identity perfectly consistent from first frame to last: cropped black hair tucked under a teal knit cap, a small crescent-shaped scar above her right eyebrow, round copper diving goggles, a faded marigold wetsuit with one torn seam at the left knee, and a mesh satchel with exactly three oyster shells bumping against her hip; use a gentle handheld follow camera tracking alongside her at matching speed with no cuts, while soft greenish noon light filters through suspended particles and the mood feels calm but focused, 16:9
OpenAI: GPT-6 Luna: Model A shows the diver’s full body swimming steadily over the lavender meadow, with a more visible kick and round copper-toned goggles; Model B has stronger realism and a clearer mesh satchel, but its tight crop obscures much of the swimming action and neither model matches the marigold wetsuit or all specified identity details. Model A’s character and camera framing remain fairly consistent across the sampled frames, while Model B’s close framing limits how well the requested shot can be judge | Anthropic: Claude Fable 5.1: Model A keeps a full-body view with a hip satch…
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 A shows a clear, continuous change in viewpoint around the steaming bowl while keeping it prominent under warm restaurant lighting. Model B is convincingly realistic, but its framing and viewpoint change only slightly, so it falls short of the requested roughly 180-degree orbit. (Second judge pass, order swapped — scores are the average of both: Model B has a warm, convincing ramen scene, but the movement reads mostly as a push-in rather than the requested orbit. Model A keeps the steaming | Anthropic: Claude Fable 5.1: Model A clearly executes a substantial orbit ar…
Matchup powered by OpenRouter.