Head to head: AuraFlow vs Luma Uni-1 Edit

AuraFlow’s problem here is simple: it keeps giving you a good-looking approximation when the assignment calls for exact editorial obedience. Luma Uni-1 Edit won the aggregate 25.3 to 17.8, and it earned that margin by being the model that consistently put the right objects, spatial relationships, and mood on the page.

The shellfish-grader test is the clearest example. Luma’s image actually reads like the requested documentary scene: weathered cedar deck, golden-hour slatted shadows, bay mist, a dented orange forklift, even hatchery text that looks readable. AuraFlow had appealing light and texture, but it drifted off brief in ways that matter — wrong aspect ratio, a kneeling pose instead of the specified waist-height three-quarter documentary framing, and equipment that felt less convincingly part of a working hatchery.

The Heliotrope Ledger palette task exposed an even bigger gap in discipline. Luma delivered the mezzanine office above the robotics repair shop, placed one accountant at the open floor safe and another marking the tall invoice board, and stayed far closer to the six-color palette. AuraFlow looked slick, but slick was the wrong priority: it missed the mezzanine-above-shop setup and leaked obvious off-palette blues and metallic neutrals into a prompt that was explicitly about controlled color and diagrammatic clarity.

Then in the rainy bus-depot mechanic scene, Luma again looked more like a photograph made in a real place. The mixed warm/cool lighting, wet concrete reflections, material realism, and mechanic-plus-coworker diagnostic setup all tracked the prompt better. AuraFlow’s version was competent, but it felt staged, skipped the opened wheel hub detail, and still couldn’t make signage and text feel fully believable.

Final call: Luma Uni-1 Edit wins, comfortably. AuraFlow is capable of handsome images, but in this head-to-head it repeatedly treated the prompt as a suggestion. Luma treated it as instructions — and that’s why it’s the better image model here.

How they were tested

We ran 3 fresh image tasks, generated on the fly for this matchup so neither model could prepare in advance, and had gpt-5.4 score each one. AuraFlow scored 17.8 to Luma Uni-1 Edit's 25.3.

1. Golden-hour shellfish grader

Documentary-style editorial photograph, 16:9 — on the weathered cedar sorting deck of the tiny shellfish hatchery 'Brackett Inlet 14', a woman in lime-green chest waders calibrates a handheld refractometer while trays of violet-lipped scallop seed sit in neat rows beside a dented orange forklift; low golden-hour sunlight from the left throws long slatted shadows through the railing, backlights mist from the bay, rims her loose hair and the wet metal tools with warm amber glow, and sets a quiet end-of-shift mood; three-quarter view at waist height, crisp realistic detail, no other time-of-day cues.

Winner: Luma Uni-1 Edit — Image B better matches the documentary editorial brief with a weathered cedar deck, visible golden-hour slatted shadows, bay mist, a dented orange forklift, and even readable hatchery text; although the handheld device is less clearly a refractometer and the waders are imperfect, it captures more of the specified scene. Image A has strong lighting and detail, but it is the wrong aspect ratio, shows a kneeling pose rather than a waist-height three-quarter documentary view, and the device/trays feel less convincingly integrated into the hatchery setting.

2. Strict Heliotrope Ledger palette

Isometric 3D render of two forensic accountants at work in a compact mezzanine office above a robotics repair shop, one crouched by an open floor safe and one highlighting discrepancies on a tall invoice board, composed in clean axonometric view, 16:9; the entire image must use only the named 'Heliotrope Ledger' palette with no stray colors whatsoever: deep aubergine, dusty mauve, heliotrope purple, parchment cream, oxidized brass, and charcoal black; every object, skin tone, paper, screen glow, shadow, and reflection must be mapped strictly into those six palette colors, with smooth controlled shading and zero off-palette reds, blues, greens, or neutrals.

Winner: Luma Uni-1 Edit — Model B matches the prompt more closely with a clear mezzanine office above a robotics repair shop, one accountant at an open floor safe and one marking a tall invoice board in clean axonometric view, while staying much closer to the six-color Heliotrope Ledger palette. Model A is polished but misses the mezzanine-above-shop setup and introduces obvious off-palette blues and metallic neutrals.

3. Rainy bus-depot mechanic realism

Photorealistic cinematic scene, 16:9 — inside the maintenance bay of municipal electric bus depot 'Route 88 Annex', a transit mechanic in a navy quilted jacket kneels beside an opened wheel hub of a teal bus while a coworker on a rolling stool checks a diagnostic tablet; overhead sodium-vapor fixtures mix with cool daylight spilling through a half-open corrugated door after rain, creating believable color contrast, soft bounce light, sharp specular highlights on wet concrete and brushed steel tools, dull grease on rubber seals, matte paint on the bus body, frayed canvas gloves, realistic skin texture, and physically accurate shadows and reflections; medium-wide eye-level composition with convincing material separation and no stylization.

Winner: Luma Uni-1 Edit — Image B better matches the prompt’s photorealistic depot-bay mood, with convincing warm/cool mixed lighting, wet concrete reflections, realistic materials, and a clearer mechanic-plus-coworker diagnostic setup. Image A is solid but feels more staged, misses the opened wheel hub detail, and its text rendering/signage is less convincing.

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