Head to head: AnimateDiff vs Happy Horse 1.1 Image to Video

On the aggregate, Happy Horse 1.1 Image to Video wins decisively: 34.3 to 17.2, with a 97% confidence verdict and a clean 4-0 sweep in task wins. That matters, because this wasn’t a case of one model edging out the other on taste. Happy Horse was simply more reliable at turning prompts into the specific shots those prompts asked for.

The clearest pattern was prompt obedience under motion. In the bakery receipt-run test, Happy Horse actually staged the requested sequence: a woman in a green apron grabs the receipt and moves through the bakery in a continuous handheld-feeling shot. AnimateDiff, by contrast, wandered into a mostly static bread-handling scene with the wrong subject and missed the core action entirely. The same story showed up in the dye-vat task, where Happy Horse better preserved the cobalt tube, stainless vat, orange gloves, magenta lighting, and the pivot toward the drying rack.

Happy Horse also had the edge when the prompt depended on temporal change being legible. In the lighting-transition scene, it more clearly sold the room shifting from warm sunset tones to cool blue dusk, with the lamp turning on in a believable locked-off shot. AnimateDiff’s version was pleasant to look at, but the key progression was weaker. In fluid dynamics, Happy Horse again read closer to the assignment: a slow-motion water-balloon burst with clearer splash evolution and droplets, while AnimateDiff veered toward something more abstract and membrane-like.

One caveat is impossible to ignore: the order-swapped judge notes point in the opposite direction on every task, which suggests some evaluation instability at the per-example explanation level. But the top-line result still lands hard: aggregate score, task count, and confidence all line up in Happy Horse’s favor. When the scoring says 4-0 and 97% confidence, the editorial takeaway is straightforward.

Final call: Happy Horse 1.1 Image to Video is the stronger video model here, and not by a hair. It was the better shot-maker, the better prompt follower, and the more dependable choice across all four tests.

How they were tested

We ran 4 fresh video tasks, generated on the fly for this matchup so neither model could prepare in advance, and had gpt-5.4 score each one. AnimateDiff scored 17.2 to Happy Horse 1.1 Image to Video's 34.4.

1. Lighting transition

A living room at dusk as the sun sets over a few seconds: the warm light visibly dims and shifts to cool blue while a lamp flickers on, the transition smooth and believable, locked-off frame, 16:9.

Winner: Happy Horse 1.1 Image to Video — Model B matches the prompt much better: the room clearly transitions from warm sunset light to cool blue dusk and the lamp turns on smoothly in a believable locked-off shot. Model A is aesthetically pleasing, but the lighting change is weaker and less clearly shows the requested warm-to-cool transition with lamp flicker-on behavior. (Order-swapped judge pass: Model A closely matches the prompt with a clear dusk-to-cool-blue transition and a believable lamp turning on in a locked-off 16:9 living room shot. Model B is aesthetically pleasant but misses the requested lighting progression and lamp flicker, with little visible temporal change across frames.)

2. Bakery Receipt Run

Single continuous shot — One unbroken take inside a neighborhood bakery at 5:42 a.m.: a sleep-deprived owner in a flour-smudged green apron hurries from the rattling receipt printer at the front counter to the proofing rack in back, snatching up a curling ticket for a 96-mini-bun catering order while calling instructions, then hooks around a sack of rye flour, nudges a swinging half-door open with her elbow, and reaches the stainless prep table as an assistant slides trays into frame; the camera starts low on the printer spitting paper, then backs ahead of her in a slightly breathless handheld retreat through the narrow aisle, panning and drifting to keep her centered without any cut, transition, or time jump; lit by warm under-cabinet bulbs mixed with pale dawn leaking through the storefront, the mood is urgent but practiced, 16:9

Winner: Happy Horse 1.1 Image to Video — Model B clearly follows the prompt with a woman in a green apron grabbing a receipt and moving through the bakery in a continuous handheld-feeling shot toward the prep area, while Model A shows a mostly static male baker handling bread and misses the key action beats. B also has stronger camera motion and scene progression, though not every specified detail is evident. (Order-swapped judge pass: Model A clearly matches the requested bakery receipt-run scenario with a woman in a green apron moving from printer through the bakery to a prep table in what appears to be a continuous handheld shot. Model B is visually stable but depicts a different static bread-handling scene with the wrong subject, no receipt-run action, and little camera movement.)

3. 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.

Winner: Happy Horse 1.1 Image to Video — Model B better matches a convincing slow-motion water-balloon burst, with clearer splash evolution, scattered droplets, and stronger side-lit cinematic presentation. Model A is visually elegant but behaves more like a surreal floating membrane and sphere than a bursting water balloon, reducing adherence and motion realism. (Order-swapped judge pass: Model A better matches a bursting water balloon in slow motion, with a convincing expanding water sheet, scattered droplets, dark background, and strong side lighting. Model B is visually attractive but reads more like an abstract translucent orb with fabric-like ribbons than a water balloon burst, reducing prompt adherence and physical plausibility.)

4. Dye Vat Turbulence

Fluid & particle dynamics — In a neon-sign workshop, a technician in orange nitrile gloves hauls a 27-centimeter cobalt glass tube out of a waist-high stainless dye-cooling vat, sending indigo water streaming off the tube and spawning overlapping ripples, droplets, tiny bubbles, and a drifting mist of steam that all behave naturally as she pivots toward a drying rack; the camera makes an energetic handheld documentary sidestep from her left hip to over her shoulder, staying close as the sloshing surface, wavering reflections, and falling beads remain physically convincing; lit by flickering magenta test lamps and a cold skylight, the mood is tense, industrious, and humid, 16:9

Winner: Happy Horse 1.1 Image to Video — Model B matches the prompt far better with a cobalt glass tube, stainless dye vat, orange gloves, magenta lighting, and a plausible pivot toward a drying rack; its water ripples and reflections also read more naturally. Model A is visually moody but misses key prompt elements and shows less coherent action around the tube and vat interaction. (Order-swapped judge pass: Model A closely matches the neon-sign workshop setup, cobalt glass tube, orange gloves, stainless vat, magenta lighting, and the pivot toward a drying rack, with believable ripples and runoff. Model B is visually striking but misses the prompt badly in subject, setting, aspect ratio, and action, showing a different worker and vat interaction unrelated to hauling a glass tube from dye coolant.)

See every prompt and the full side-by-side outputs in the interactive Head-to-Head.