Head to head: Gemini Omni Flash 1.1 Image to Video vs LTX-2.3 22B Distilled
Gemini Omni Flash 1.1 Image to Video vs LTX-2.3 22B Distilled
By Ryan Merket · Published
This matchup tests whether a heavyweight distilled video model can keep pace with a faster multimodal rival across controlled camera moves, athletic action, lighting changes, and character continuity. The result turns on prompt fidelity and temporal direction, not merely attractive individual frames.
Gemini Omni Flash 1.1 Image to Video dominates this matchup: **32.0 to 12.8**, with a **12–0 sweep across tasks** and a clear win at **95% statistical confidence**. That margin reflects the same advantage repeated across every tested scenario, not one spectacular clip skewing the aggregate. The sharpest separation is in directed motion and sequencing. Gemini produces recognizable, steaming ramen while sustaining a largely level orbit; LTX often substitutes vegetable soup and turns the move into a push-in or steep reframing. In the hammer-toss tests, Gemini preserves the dusk field, throwing ring, implement flight, rotation, and recovery. LTX repeatedly loses the requested setting, equipment scale, or athletic action altogether. Gemini is equally stronger when the prompt requires events to happen in a specific order. Its living rooms move convincingly from warm sunset to cool dusk before the lamp becomes prominent, while LTX commonly leaves the lamp on throughout or barely changes the lighting. In the rink tunnel, Gemini maintains the frontal, backward-tracking skater portrait and shifts from green fluorescence to magenta sponsor light; LTX tends toward rear-facing or animated compositions, missing the facial performance and neon activation. Gemini is not flawless: ramen toppings drift, sponsor lettering mutates, and the chalk burst or neon sputter can be understated. But those are execution blemishes inside clips that still understand the assignment. LTX’s failures are more fundamental—wrong subjects, wrong camera paths, missing actions, and weaker temporal staging. **Final call: Gemini Omni Flash 1.1 Image to Video wins decisively. The 12–0 sweep and limited confidence make this a clear verdict, with materially better prompt adherence, camera control, realism, and event sequencing.**
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 clearly presents steaming ramen under warm restaurant lighting and shows a coherent circling move with largely stable height and framing. Model B maintains reasonable visual continuity, but the dish does not read as ramen, lacks visible steam, and the substantial push-in undermines the requested constant orbit. (Second judge pass, order swapped — scores are the average of both: Model A clearly depicts recognizable steaming ramen under warm restaurant lighting, with a coherent orbit and s | Anthropic: Claude Fable 5.1: Model A depicts a convincing steaming rame…
Hammer toss chalk burst
One continuous 16:9 shot at dusk on a quiet practice field: a masters hammer throw athlete in a faded maroon singlet completes a full release inside the ring, and the camera starts in a low intimate close-up near his lead hip then arcs sideways and slightly backward to follow the first part of the hammer’s flight while keeping his body, the wire, and the toe board in frame; as he pivots, chalk dust from his shoes puffs and trails across the wet concrete, the heavy ball pulls his torso off-center before release, his jersey ripples from the rotational speed, and after the hammer leaves his hands his momentum carries him into a believable recovery step against the cage line; stadium lamps have just clicked on, adding cool highlights over the last blue-orange twilight, with a tense, disciplined training-ground mood, no cuts.
OpenAI: GPT-5.6 Sol Pro: Model A convincingly depicts a dusk hammer release on a wet throwing field with visible cage, wire, lamps, rotational movement, and recovery, though the chalk burst and intimate tracking arc are not clearly demonstrated. Model B features an older athlete and wet ground but has implausibly oversized equipment, weak field/ring fidelity, awkward motion, and no convincing hammer flight. (Second judge pass, order swapped — scores are the average of both: Model A clearly depicts a coherent hammer rele | Anthropic: Claude Fable 5.1: Model A delivers a coherent dusk practice…
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 clearly depicts a locked-off living room transitioning from warm sunset light to cool dusk, followed by the lamp switching on, with strong realism and consistency. Model B remains largely unchanged in lighting, has lamps already illuminated, and focuses on a walking figure rather than the requested lighting transition. (Second judge pass, order swapped — scores are the average of both: Model A clearly depicts a locked-off, believable progression from warm sunset light to cool dusk follow | Anthropic: Claude Fable 5.1: Model A nails the prompt with a locked-off…
Flickering rink tunnel
One continuous 16:9 shot in the service tunnel beneath a small-city ice arena: a teenage speed skater in a cobalt skinsuit rolls slowly toward the rink door on blade guards, flexing one gloved hand and breathing into the cold air, while the camera makes an intimate close-up tracking move backward at chest height, keeping her face and shoulders centered; halfway through, a row of magenta-pink sponsor neon letters above the concrete wall sputters on in uneven pulses and then stabilizes, washing her cheekbones, visor, and the damp tunnel floor with changing colored light as the earlier greenish fluorescent ambience fades under it; the reflections should shift smoothly across the helmet and cinderblock wall, with a focused, pre-race, slightly nervous mood, no cuts.
OpenAI: GPT-5.6 Sol Pro: Model A closely matches the intimate backward-tracking tunnel shot, maintains a consistent skater, and clearly transitions from green fluorescent ambience to reflected magenta sponsor neon. Model B has striking color but begins behind the skater, lacks the requested sponsor-letter activation, and shows a jarring viewpoint and facial-style change. (Second judge pass, order swapped — scores are the average of both: Model A closely matches the centered backward-tracking portrait, realistic arena tu | Anthropic: Claude Fable 5.1: Model A delivers a photoreal teenage skate…
Matchup powered by OpenRouter.