Head to head: Boreal vs Luma Ray 2 (Image to Video)
Boreal vs Luma Ray 2 (Image to Video)
By Ryan Merket · Published
This matchup turns on controllability rather than surface polish: lighting changes, camera orbits, occlusion logic, and wind-driven water physics separate the two models. Both produce convincing moments, but their failure modes are sharply different.
Boreal wins the benchmark 27.8 to 21.7, taking seven of 12 tasks with no ties. The limited confidence result makes this a clear statistical victory, not a narrow call disguised by the aggregate score. The decisive advantage was lighting control. Boreal swept all three lighting-transition tests, consistently preserving a locked-off living room while moving from warm sunset to cool dusk and switching on the lamp naturally. It also won two of three camera-motion trials because its viewpoint changes actually read as substantial orbits; Luma Ray 2 often looked cleaner and more stable, but moved too little to satisfy the requested roughly 180-degree arc. Luma Ray 2 was strongest when local continuity and physical texture mattered. It won two of three gust-driven bucket-pour tests, producing more weighted water, wind-broken droplets, impact ripples, and settling behavior, while Boreal repeatedly turned the pour into an oversized, implausibly smooth waterfall. Luma also took two of three cyclist trials, benefiting from steadier rider and bicycle continuity—even though it often failed to execute the prompt’s crucial foreground-gauge occlusion. That contrast defines the matchup: Boreal more reliably performs the requested temporal transformation or camera instruction, but can deform objects and mishandle fluid scale; Luma Ray 2 can deliver stronger photorealism and short-range consistency, yet too often sidesteps the central action. **Final call: Boreal wins clearly—less polished at times, but substantially more dependable when the prompt requires deliberate change over time.**
Hail cyclist occlusion
A short, continuous 16:9 clip in a mountain weather station yard during a sudden hail squall: a woman in a mustard rain cape pedals a dented teal cargo bicycle from left to right past stacked white instrument crates, while the handheld camera hustles sideways to track her at waist height; she disappears fully for a moment behind a large spinning silver precipitation gauge and a flapping orange tarp, then re-emerges unchanged on the far side with the same cape, bike, and posture, still moving smoothly as hail rattles on metal, cold blue-gray overcast light reflecting off wet concrete, tense documentary mood, no cuts.
OpenAI: GPT-5.6 Sol Pro: Model A better depicts the specified silver gauge occlusion, orange tarp, crates, and cold industrial weather-station setting, though the cyclist’s orientation and bicycle geometry shift noticeably across the occlusion. Model B maintains a more consistent cyclist and rightward trajectory, but omits the crucial gauge-driven full occlusion and shows little visible hail. (Second judge pass, order swapped — scores are the average of both: Model A better realizes the central precipitation-gauge and o | Anthropic: Claude Fable 5.1: Model A includes the key occlusion setup—t…
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.
OpenAI: GPT-5.6 Sol Pro: Model A convincingly shows a locked-off living room progressing from warm sunset to cool dusk as the lamp illuminates, with strong scene continuity and believable lighting. Model B begins on a patio rather than in a living room, shows little transition, and abruptly changes to an unrelated interior with subtitles, severely breaking temporal consistency. (Second judge pass, order swapped — scores are the average of both: Model B abruptly changes from an outdoor terrace to an unrelated interior wi | Anthropic: Claude Fable 5.1: Model A delivers exactly the prompt: a loc…
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-5.6 Sol Pro: Model A shows a clearer circling move at a fairly stable height, though the toppings noticeably morph and the sampled views do not convincingly span 180 degrees. Model B is visually cleaner and more temporally consistent, but its framing changes too little to read as the requested orbit. (Second judge pass, order swapped — scores are the average of both: Model A shows a clearer continuous orbit around the bowl with warm restaurant lighting and stronger composition, though the noodles, greens, an | Anthropic: Claude Fable 5.1: Model A actually performs a visible orbit…
Gust bucket pour physics
A short, continuous 16:9 clip on a windswept coastal bluff at golden hour just before a storm: a park ranger in a faded burgundy field jacket stands beside a wooden rain-measurement post and tips a galvanized bucket of seawater into a narrow rock tide pool, while the camera starts close on the bucket and arcs down and right in one energetic handheld move to follow the stream; the water must pour with believable weight, break into sheets and droplets in the crosswind, splash against the stone, rebound, and settle into ripples under gravity and momentum as the ranger braces against gusts, warm low sun cutting through dark storm clouds, dramatic but natural mood, no cuts.
OpenAI: GPT-5.6 Sol Pro: Model B better matches the narrow tide pool, storm-lit coastal setting, ranger, and galvanized-bucket action, with convincing impact ripples and stronger overall composition. Model A follows the falling stream effectively, but the oversized waterfall-like sheet and open rocky basin weaken the requested crosswind physics and tide-pool specificity. (Second judge pass, order swapped — scores are the average of both: Model B better matches the requested close-starting composition, narrow tide pool, | Anthropic: Claude Fable 5.1: Model A includes the ranger, wooden post an…
Matchup powered by OpenRouter.