Head to head: H3 Max Hand Drawn vs Kling Video V3 Turbo Pro Image to Video
H3 Max Hand Drawn vs Kling Video V3 Turbo Pro Image to Video
By Ryan Merket · Published · Updated
H3 Max Hand Drawn and Kling Video V3 Turbo Pro Image to Video were tested on camera movement, scene choreography, character continuity, and latte-art action. The comparison exposes a clear difference in style, but no dependable overall gap in performance.
The aggregate scores are 27.1 for H3 Max and 27.9 for Kling—a **0.8-point margin**, with only limited confidence that either model is genuinely better. That is a sample tie, not a narrow win to dress up as a verdict. Across the repeated tasks, each model had moments where it better matched the prompt, and the results shifted between runs. H3 Max’s strongest case is camera control. In the ramen tests, it repeatedly made the viewpoint travel around the bowl, while Kling’s more photographic clips often held nearly the same framing. H3 also sometimes made the latte pour read more clearly as a rosetta, with a stronger overhead composition. The tradeoff: its hand-drawn rendering can pull away from prompts asking for naturalistic video. Kling’s advantage is often the finished live-action look, and it sometimes paired that with better prompt fit: it captured key Lisbon setting details in some tram-stop runs and produced convincing courier or latte-art results in others. But that consistency wasn’t universal. Its camera movement could be too slight, and some clips missed requested actions such as the courier’s hop or a clean overhead rosetta. **Final call: tie.** H3 Max is the more reliable choice for visible camera motion; Kling is often the more convincing realistic scene. Neither advantage holds across the full set, and the score gap is too small to justify naming an overall winner.
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 change in viewing angle around the steaming ramen, making it much closer to the requested continuous orbit. Model B is more photorealistic and stable, but its framing barely changes across the sampled frames, so it does not convey the requested camera motion. (Second judge pass, order swapped — scores are the average of both: Model A shows a clear, continuous change in viewpoint around the steaming bowl, while Model B remains nearly static across the sampled frames. Model B | Anthropic: Claude Fable 5.1: Model A clearly executes the requested ~180° or…
Tram stop weather ballet
A short continuous 16:9 video clip at a rainy hillside tram stop in Lisbon just after sunrise, alive with layered ambient motion: low clouds drift between apartment blocks, plane trees sway and drip, puddles on the cobblestones ripple from intermittent drops, steam curls steadily from a chestnut cart, and a torn orange timetable flutters against the shelter wall; in the foreground, a middle-aged man in a navy mechanic’s coverall carries two stacked pastry boxes, pauses to shift his grip, then steps around a sleeping tan dog as the camera makes a fluid gliding gimbal arc from his side to slightly ahead of him, revealing the street descending toward the river, lit by pale gold morning light breaking through wet gray air, with a calm, observant, lived-in city mood.
OpenAI: GPT-6 Luna: Model A more fully conveys the rainy, atmospheric prompt, with the steaming cart, torn orange timetable, wet street, dog, and a clear change in the man’s movement as the view glides toward the river. Model B is sharper and more photorealistic, but its scene feels more static and omits much of the requested ambient weather motion. (Second judge pass, order swapped — scores are the average of both: Model B more closely matches the requested realistic Lisbon scene, with the mechanic, pastry boxes, | Anthropic: Claude Fable 5.1: Model A delivers more of the prompted ambient de…
Consistent florist courier
A short continuous 16:9 video clip of the same young florist courier jogging down a narrow side street in Osaka at blue hour, never changing identity or appearance from first frame to last: she has a sharp asymmetrical bob dyed ink-black with one copper streak over her left temple, a faint crescent scar under her right eye, round moss-green glasses, a mustard windbreaker with a hand-sewn patch reading "Midori 48," charcoal bike shorts over opaque plum tights, and a square teal delivery backpack stuffed with wrapped ranunculus stems; she slows to check a paper receipt, smiles, then hops over a shallow puddle and keeps moving, while the camera performs a smooth gliding gimbal move backward in front of her, slightly lowering as it tracks, neon shop signs reflecting on wet pavement under cool streetlights and a warm ramen-window glow, with a focused, intimate, quietly determined mood.
OpenAI: GPT-6 Luna: Model A better conveys the specified bob with copper streak, plum tights, flower-filled backpack, and the continued running movement, though its illustrated look departs from a naturalistic video style. Model B has polished live-action visuals and clearly shows the receipt check and smile, but the sampled frames do not show the requested hop or continued jogging and omit more of the distinctive appearance details. (Second judge pass, order swapped — scores are the average of both: Model B has po | Anthropic: Claude Fable 5.1: Model A nails nearly every character detail (co…
Subject action
A barista's hands pouring latte art: the milk stream forms a clean rosetta in the crema with natural, fluid wrist motion, no cuts, overhead close-up, soft café light, 16:9.
OpenAI: GPT-6 Luna: Model A shows a clear, coherent progression from pouring to a clean rosetta in an overhead close-up, though its illustrated style is less realistic. Model B is more photographic and the pour looks fluid, but the dark shape in the first frame and the less tidy final pattern weaken its consistency and prompt fit. (Second judge pass, order swapped — scores are the average of both: Model B has a realistic overhead café close-up and develops a recognizable rosetta, though the initial pour is messy. M | Anthropic: Claude Fable 5.1: Model B delivers a photorealistic overhead pour…
Matchup powered by OpenRouter.