Head to head: Happy Horse 1.1 Image to Video vs Heygen
Happy Horse 1.1 Image to Video vs Heygen
By Ryan Merket · Published · Updated
A close image-to-video comparison across pouring, camera movement, fluid effects and handwork, judged on a limited set of saved tasks. The key question is which model more consistently makes the requested action and framing legible.
Happy Horse 1.1 Image to Video repeatedly made the prompt’s defining motion easier to read: milk pouring into a rosetta, a centered slow dolly-in, and a balloon’s clear burst into water and droplets. Heygen often produced attractive imagery, but in these comparisons it more often missed a requested action or shifted the framing—most notably skipping the pour in one rosetta clip and drifting toward an overhead angle in the dolly-in tasks. The dumpling results were mixed: Heygen won one task with clearer fingertip pleating, while Happy Horse won the other by keeping the handwork closer and more distinct. Across the saved tasks, Happy Horse had 7 task wins to Heygen’s 1, with 0 ties; aggregate scores were 31.2 and 25.0, respectively. Those figures describe the task scorecard, but the saved overall result remains a tie. **Editorial strength: Too close to call (not statistical).** Two successful repeated trials yielded a mean paired score difference (A−B) of 6.25, with a reported Student-t 95% confidence interval of -9.6327559205 to 22.1327559205. That interval is not a probability of winning or of being generally better. This assessment applies only to the saved task sample. **Final call: Tie.** Happy Horse had the broader edge on the tested actions, but the saved result does not declare an overall winner.
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 clearly shows the milk being poured as a clean rosetta develops, with a steady overhead composition and fluid hand positioning. Model B has an attractive finished rosetta, but it skips the pouring action and shifts abruptly from a partial cup to the completed drink. (Second judge pass, order swapped — scores are the average of both: Model B shows an attractive finished rosetta, but the sampled frames do not show the milk being poured and jump from a partially cropped cup to the completed | Anthropic: Claude Fable 5.1: Model A actually shows the barista's hands pour…
Camera motion control — slow dolly-in
A short one-shot video in a cramped midnight noodle bar pantry: a tattooed prep cook in a faded teal apron rapidly brushes chili oil over twelve skewers of king oyster mushrooms arranged like clock spokes on a dented aluminum tray, while steam drifts from a stockpot in the background; the camera performs a smooth, exact slow dolly-in from waist-up framing to a tight close view of the glossy mushrooms and moving hands, with no lateral drift or tilt, preserving the subject centered throughout; lit by a single warm fluorescent strip overhead with cool spill from a refrigerator door, creating a tense, focused, after-hours mood, 16:9
OpenAI: GPT-6 Luna: Model A delivers a clear, centered slow dolly-in from waist-up framing to a close view of the brushing hands and glossy spoke-arranged mushrooms, while maintaining the moody pantry lighting and consistent perspective. Model B has appealing close food detail, but starts much tighter and shifts toward a more overhead angle, weakening the requested framing and no-tilt camera motion. (Second judge pass, order swapped — scores are the average of both: Model A better preserves the requested waist-up-t | Anthropic: Claude Fable 5.1: Model A delivers a clean, centered slow dolly-i…
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.
OpenAI: GPT-6 Luna: Model A clearly shows a red balloon bursting into an expanding water sheet and scattering droplets, with strong side lighting against a dark background. Model B has striking fluid detail, but the balloon and burst are less legible, and the motion reads more like a clear orb collapsing than a water balloon bursting. (Second judge pass, order swapped — scores are the average of both: Model A clearly shows a suspended balloon bursting into a broad water sheet and scattering droplets, with crisp dar | Anthropic: Claude Fable 5.1: Model A clearly depicts a red water balloon on…
Subject action & gesture — pleating dumplings
A short continuous shot in a small riverside test kitchen at dawn: an elderly chef with silver braids and flour on her knuckles stands at a wooden counter and completes three distinct soup dumplings in sequence, scooping orange crab filling, folding each wrapper, and pleating the rims shut with quick, precise fingertip pinches and natural wrist turns before setting them into a bamboo steamer; the camera makes a gentle handheld follow from her right side, staying close to her hands and torso without cuts as her body shifts subtly with the work; soft blue morning window light mixes with a dim amber task lamp, giving the scene a quiet, intimate, concentrated mood, 16:9
OpenAI: GPT-6 Luna: Model A establishes the chef and orange filling, but the sampled action barely shows distinct fingertip pleating. Model B stays close on the hands and clearly depicts the dumpling being pinched into pleats, with finished dumplings in the steamer, though it omits the filling and broader chef details. (Second judge pass, order swapped — scores are the average of both: Model B makes the prompt’s defining action especially legible: precise fingertip pleats form as finished dumplings accumulate in th | Anthropic: Claude Fable 5.1: Model B keeps the camera tight on the chef's ha…
Matchup powered by OpenRouter.