Head to head: Luma Ray 3.2 Image to Video vs PixVerse V4 Image To Video
Luma Ray 3.2 Image to Video vs PixVerse V4 Image To Video
By Ryan Merket · Published
Two image-to-video models take on demanding camera moves, lighting changes, and athletic action. The comparison turns on which one follows the brief more reliably—not just which clips look polished at a glance.
Luma Ray 3.2 takes the matchup, 26.4 to 24.4, with a tentative lean verdict. It won seven tasks to PixVerse V4’s three, with two ties: a clear overall result, but not a blowout. The difference is consistency in executing the requested scene and camera behavior, rather than a universal edge in image quality. Luma’s strongest showing came in the cathedral prompts. It repeatedly delivered a grand nave, centered forward movement toward the altar, and warm, dusty candlelight; PixVerse’s versions often looked polished but smaller in scale, more like a chapel, or less committed to the glide. Luma also generally handled the velodrome brief better, with low tracking action, damp boards, and a shift toward amber reflections—even when crops or lighting changes fell short. PixVerse made its clearest case on the river-stair sprint. Its runner, baton, marina setting, and side-to-front follow more often matched the requested orbit, while Luma sometimes obscured the baton, favored a frontal view, or introduced an unwanted camera operator. That is a real, repeatable strength, not a consolation prize. Lighting transitions were the messiest category for both. Luma usually came closer to the requested dusk shift, but some clips barely changed or drifted from the locked-off framing; PixVerse could hold a stable composition, yet sometimes missed the transition outright—or veered into an entirely wrong effect. **Final call: Luma Ray 3.2 wins on more dependable prompt-following and stronger cinematic continuity; choose PixVerse when the moving runner-and-camera relationship is the priority.**
Floodlights awaken over velodrome sprinter
A single continuous 5-second shot at the rain-damp Marrowbank Velodrome: a female track sprinter in a matte teal skinsuit explodes out of turn three and drives down the straight while the camera glides low alongside her in a smooth lateral tracking move, matching her speed and keeping her waist-up centered; halfway through, the arena’s sodium floodlights thrum on in staggered banks after a dim overcast lull, and the scene transitions visibly from cool gray ambient light to warm amber reflections racing across the wet boards and her chrome crankset, with a tense, electric pre-race mood, 16:9
OpenAI: GPT-6 Luna: Model A better matches the low lateral tracking view, centered teal-clad sprinter, wet boards, and visible shift toward warm amber reflections. Model B is smooth and polished, but follows the rider from behind, keeps the arena cool-lit, and misses the requested waist-up side view. (Second judge pass, order swapped — scores are the average of both: Model A better matches the low lateral tracking shot and shows a forceful sprint on wet boards, with the lighting warming over the sequence. Model B h | Anthropic: Claude Fable 5.1: Model A delivers a low lateral tracking shot al…
Orbit-follow on river stair sprint
One continuous 6-second shot on the slate steps beside the Vesper Rowing Basin: a teenage decathlete in a scarlet singlet sprints diagonally upward two steps at a time carrying a sand-filled training baton, and the camera performs a dramatic handheld-feel orbiting follow from slightly below his right side to just ahead of him, continuously adjusting to keep his face and torso sharp and framed as the stone balustrade, rippling water, and blurred oarsheds slide behind him; late blue-hour light with cold sky fill and a row of pale marina lamps reflecting in the river creates a gritty, determined mood, 16:9
OpenAI: GPT-6 Luna: Model B more clearly shows a baton-carrying runner sprinting up the steps and its framing shifts from a side-follow to a more frontal angle against the marina at blue hour. Model A better suggests a teenager, but the baton is unclear and a camera operator intrudes in the final frame, weakening the continuous-shot illusion. (Second judge pass, order swapped — scores are the average of both: Model B more clearly shows a baton-carrying sprinter climbing the slate steps, with a convincing side-to-fr | Anthropic: Claude Fable 5.1: Model B delivers a clean, well-lit blue-hour ma…
Single continuous shot
One unbroken take gliding slowly through a candlelit cathedral from the entrance toward the altar, no cuts, jumps, or transitions, dust and warm light in the air, 16:9.
OpenAI: GPT-6 Luna: Model A better conveys a vast cathedral entrance-to-altar glide, with a stable centered aisle, atmospheric dust, and warm candlelight across the sampled frames. Model B has a clear forward movement and polished imagery, but feels more like a compact chapel and shows less of the requested hazy atmosphere. (Second judge pass, order swapped — scores are the average of both: Model A better conveys a vast cathedral with warm, dusty candlelit atmosphere, and its successive frames show a steady glide t | Anthropic: Claude Fable 5.1: Model A delivers a grand candlelit cathedral wi…
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-6 Luna: Model A shows a more noticeable dusk progression and keeps the framing steadier, though the lamp is already lit and the lighting shift is subtle. Model B is visually polished, but its lighting stays nearly unchanged and the framing shifts slightly by the final frame. (Second judge pass, order swapped — scores are the average of both: Model A keeps a stable locked-off composition and shows a subtle dusk shift, while Model B has stronger color and visual impact but its framing drifts. Both show th | Anthropic: Claude Fable 5.1: Neither clip delivers a clear warm-to-cool dimm…
Matchup powered by OpenRouter.