{ "date": "2026-09-21", "method": "GPT Image 2.5 Sunburst (low) replaces GPT Image 2 in the six-model comparison. Scores for the other five models are retained from the original blind review. Sunburst was reviewed non-blind. No combined ranks or win counts are assigned; image strips sort by score, with ties in fixed model order.", "model": "openai/gpt-image-2.5/sunburst/edit", "quality": "low", "count": 23, "summary": [ { "key": "runchat-image", "name": "Runchat Image", "credits": 4, "score": 4.956521739130435, "latencySeconds": 7.128739130434783, "dims": { "design": 6.086956521739131, "style": 5.739130434782608, "prompt": 5.391304347826087, "aesthetic": 6 }, "capabilities": { "render_single_ref": 5.5, "render_multi_ref": 4, "render_json_no_ref": 5, "site_placement": 5.5, "object_insertion_lighting": 7.5, "local_edit": 5, "novel_views": 5, "concept_merge": 5.5, "material_transfer": 3, "graphics_icons": 5, "poster_generate": 2.5, "relight_realism": 5.25 } }, { "key": "muse-image", "name": "Muse Image", "credits": 20, "score": 8.08695652173913, "latencySeconds": 37.03417391304348, "dims": { "design": 8.217391304347826, "style": 8.130434782608695, "prompt": 8, "aesthetic": 8.26086956521739 }, "capabilities": { "render_single_ref": 8.5, "render_multi_ref": 8, "render_json_no_ref": 9, "site_placement": 7.5, "object_insertion_lighting": 8, "local_edit": 6.5, "novel_views": 8, "concept_merge": 8.5, "material_transfer": 9, "graphics_icons": 7.5, "poster_generate": 8, "relight_realism": 8.5 } }, { "key": "gpt-image-2-5-sunburst-low", "name": "GPT Image 2.5 (low)", "credits": 40, "score": 7.478260869565218, "latencySeconds": 20.344565217391303, "dims": { "design": 8.043478260869565, "style": 7.782608695652174, "prompt": 7.304347826086956, "aesthetic": 7.956521739130435 }, "capabilities": { "render_single_ref": 7.5, "render_multi_ref": 6, "render_json_no_ref": 8, "site_placement": 5.5, "object_insertion_lighting": 9, "local_edit": 7, "novel_views": 8, "concept_merge": 9, "material_transfer": 9, "graphics_icons": 8, "poster_generate": 6.5, "relight_realism": 6.75 } }, { "key": "nano-banana-1", "name": "Nano Banana 1", "credits": 80, "score": 4.739130434782608, "latencySeconds": 11.190782608695653, "dims": { "design": 5.695652173913044, "style": 6.086956521739131, "prompt": 5.130434782608695, "aesthetic": 6.217391304347826 }, "capabilities": { "render_single_ref": 4.5, "render_multi_ref": 6, "render_json_no_ref": 8, "site_placement": 4, "object_insertion_lighting": 5, "local_edit": 4, "novel_views": 3, "concept_merge": 3.5, "material_transfer": 6, "graphics_icons": 4, "poster_generate": 3.5, "relight_realism": 5.75 } }, { "key": "nano-banana-pro-2k", "name": "Nano Banana Pro (2K)", "credits": 300, "score": 8.043478260869565, "latencySeconds": 33.87904347826087, "dims": { "design": 8, "style": 8.521739130434783, "prompt": 8.130434782608695, "aesthetic": 8.608695652173912 }, "capabilities": { "render_single_ref": 8, "render_multi_ref": 9, "render_json_no_ref": 7, "site_placement": 5.5, "object_insertion_lighting": 8.5, "local_edit": 8, "novel_views": 9, "concept_merge": 9, "material_transfer": 9, "graphics_icons": 9, "poster_generate": 8.5, "relight_realism": 7.25 } }, { "key": "flux-2-pro", "name": "Flux 2 Pro", "credits": 100, "score": 5.3478260869565215, "latencySeconds": 21.077565217391303, "dims": { "design": 5.695652173913044, "style": 6.391304347826087, "prompt": 5.826086956521739, "aesthetic": 6.478260869565218 }, "capabilities": { "render_single_ref": 6, "render_multi_ref": 5, "render_json_no_ref": 7, "site_placement": 4.5, "object_insertion_lighting": 6, "local_edit": 5.5, "novel_views": 4, "concept_merge": 5, "material_transfer": 2.5, "graphics_icons": 5, "poster_generate": 4.5, "relight_realism": 7.25 } } ], "tests": [ { "id": "R1-render-single-short", "capability": "render_single_ref", "prompt": "Photorealistic render of the house in image 1 using the materials and lighting of image 2.", "criteria": "View preservation is primary (the overlay row shows it), then massing fidelity, then material and light transfer, then photographic quality. SHORT prompt: measures what the model does unprompted. Weigh how closely the output matches the reference's aesthetic - light, colour, photographic style, framing - since users pick references for exactly that; and treat changes to design elements (openings, balustrades, roof form) or to the setting's character (tree species, ground cover) as faults even when the image is beautiful.", "inputs": [ "shaded_barn@3x4", "photo_house_hillside" ], "overlay": true, "cells": [ { "key": "runchat-image", "name": "Runchat Image", "credits": 4, "score": 6, "dims": { "design": 9, "style": 4, "prompt": 6, "aesthetic": 5 }, "note": "Best view hold and timber right, but sunny blue sky and green lawn - the reference is overcast and dry.", "latencyMs": 9141, "method": "original blind review" }, { "key": "muse-image", "name": "Muse Image", "credits": 20, "score": 8, "dims": { "design": 8, "style": 8, "prompt": 8, "aesthetic": 8 }, "note": "Massing, gable, wing and plinth intact, timber cladding, and the closest match to image 2's muted overcast hillside mood; slight camera drift.", "latencyMs": 21956, "method": "original blind review" }, { "key": "gpt-image-2-5-sunburst-low", "name": "GPT Image 2.5 (low)", "credits": 40, "score": 8, "dims": { "design": 8, "style": 8, "prompt": 8, "aesthetic": 8 }, "note": "Keeps the glazed gable, lower wing and raised plinth and transfers the muted hillside palette. Small framing and terrace-edge shifts remain; convincing dry planting and soft light.", "latencyMs": 21379, "method": "non-blind follow-up" }, { "key": "nano-banana-1", "name": "Nano Banana 1", "credits": 80, "score": 3, "dims": { "design": 3, "style": 6, "prompt": 4, "aesthetic": 6 }, "note": "Camera moved well off the input (double image in the blend); eucalypt setting and timber right.", "latencyMs": 11227, "method": "original blind review" }, { "key": "nano-banana-pro-2k", "name": "Nano Banana Pro (2K)", "credits": 300, "score": 8, "dims": { "design": 7, "style": 9, "prompt": 7, "aesthetic": 9 }, "note": "Photographic and tonally the closest to the reference (soft light, dry grass, muted palette), timber cladding right; added a chimney and drifted a little.", "latencyMs": 32969, "method": "original blind review" }, { "key": "flux-2-pro", "name": "Flux 2 Pro", "credits": 100, "score": 5, "dims": { "design": 5, "style": 5, "prompt": 6, "aesthetic": 7 }, "note": "Convincing photograph but it changed the design and setting: solid parapet instead of the open balustrade, different tree species, moodier sky than the reference. (Moved down on review.)", "latencyMs": 18840, "method": "original blind review" } ] }, { "id": "R2-render-single-detailed", "capability": "render_single_ref", "prompt": "Render image 1 as a photorealistic architectural photograph, keeping the exact camera position, focal length, framing and aspect ratio of image 1, with every edge of the building where it is in the frame. Keep the geometry exactly as image 1 has it: the tall gable with its fully glazed gable end, the lower flat-roofed glazed wing extending to the right, the raised plinth and the window positions. From image 2 take only materials, lighting and mood: pale timber board cladding on the solid walls, a dark standing-seam metal roof, clear glazing with slim dark frames, and the soft overcast light and dry-grass hillside setting. Ground: mown grass with a gravel path.", "criteria": "Same as R1; this is the DETAILED prompt, positive instructions only. Compare the pair: a model that only holds the view when told to scores here but not on R1. Weigh how closely the output matches the reference's aesthetic - light, colour, photographic style, framing - since users pick references for exactly that; and treat changes to design elements (openings, balustrades, roof form) or to the setting's character (tree species, ground cover) as faults even when the image is beautiful.", "inputs": [ "shaded_barn@3x4", "photo_house_hillside" ], "overlay": true, "cells": [ { "key": "runchat-image", "name": "Runchat Image", "credits": 4, "score": 5, "dims": { "design": 7, "style": 5, "prompt": 6, "aesthetic": 6 }, "note": "Best alignment score but saturated green lawn, white plinth and CG light; furthest from the reference setting.", "latencyMs": 7414, "method": "original blind review" }, { "key": "muse-image", "name": "Muse Image", "credits": 20, "score": 9, "dims": { "design": 8, "style": 9, "prompt": 9, "aesthetic": 9 }, "note": "Closest to the reference: pale boards, dark roof, dry hillside with eucalypts under soft overcast; view and massing held.", "latencyMs": 21785, "method": "original blind review" }, { "key": "gpt-image-2-5-sunburst-low", "name": "GPT Image 2.5 (low)", "credits": 40, "score": 7, "dims": { "design": 9, "style": 8, "prompt": 6, "aesthetic": 8 }, "note": "Building and camera are close to the input, with pale boards and soft overcast light. Replaces the explicitly requested mown grass and gravel path with long dry grass.", "latencyMs": 20731, "method": "non-blind follow-up" }, { "key": "nano-banana-1", "name": "Nano Banana 1", "credits": 80, "score": 6, "dims": { "design": 8, "style": 6, "prompt": 7, "aesthetic": 6 }, "note": "Design held but a bright mown lawn and flat light make it read as a render rather than the reference photograph.", "latencyMs": 11205, "method": "original blind review" }, { "key": "nano-banana-pro-2k", "name": "Nano Banana Pro (2K)", "credits": 300, "score": 8, "dims": { "design": 8, "style": 8, "prompt": 9, "aesthetic": 8 }, "note": "Very photographic, dry grass and gravel path, overcast grade like the reference; view held.", "latencyMs": 38353, "method": "original blind review" }, { "key": "flux-2-pro", "name": "Flux 2 Pro", "credits": 100, "score": 7, "dims": { "design": 8, "style": 7, "prompt": 8, "aesthetic": 7 }, "note": "View held and materials right, but a darker hill with bare trees moves the mood away from the reference.", "latencyMs": 18775, "method": "original blind review" } ] }, { "id": "R3-render-multi-short", "capability": "render_multi_ref", "prompt": "Render image 1 as a photograph. Materials from image 2, photographic style from image 3.", "criteria": "Design fidelity to the line drawing first (openings included - read the blend row), then material transfer from image 2 (brick and terracotta), then how closely the photographic style - Kodachrome colour, hazy sky, warm faded tones, slight softness, 1960s period feel - matches image 3, which is deliberately a very different aesthetic from image 2 so the two transfers can be told apart, then whether it reads as a photograph rather than a render. SHORT prompt: measures what the model does unprompted.", "inputs": [ "revit_courtyard_housing_bw@3x4", "divisare_multires1", "ref_kodachrome" ], "overlay": true, "cells": [ { "key": "runchat-image", "name": "Runchat Image", "credits": 4, "score": 4, "dims": { "design": 8, "style": 3, "prompt": 5, "aesthetic": 6 }, "note": "Clean modern render with terracotta and brick, view held, but the photographic style of image 3 is absent.", "latencyMs": 10429, "method": "original blind review" }, { "key": "muse-image", "name": "Muse Image", "credits": 20, "score": 8, "dims": { "design": 8, "style": 8, "prompt": 8, "aesthetic": 8 }, "note": "Brick and terracotta right, gentle Kodachrome warmth and haze, view held well; the most balanced result.", "latencyMs": 27415, "method": "original blind review" }, { "key": "gpt-image-2-5-sunburst-low", "name": "GPT Image 2.5 (low)", "credits": 40, "score": 6, "dims": { "design": 6, "style": 5, "prompt": 6, "aesthetic": 8 }, "note": "Transfers brick and tile but lowers the viewpoint and changes the facade proportions. The third reference's strong Kodachrome colour character is weak; reads as a conventional daylight render.", "latencyMs": 22354, "method": "non-blind follow-up" }, { "key": "nano-banana-1", "name": "Nano Banana 1", "credits": 80, "score": 6, "dims": { "design": 7, "style": 8, "prompt": 5, "aesthetic": 7 }, "note": "Good grade and materials but copied the flags and signage from image 3 wholesale, and changed the aspect ratio.", "latencyMs": 13246, "method": "original blind review" }, { "key": "nano-banana-pro-2k", "name": "Nano Banana Pro (2K)", "credits": 300, "score": 9, "dims": { "design": 7, "style": 9, "prompt": 9, "aesthetic": 9 }, "note": "Strongest Kodachrome transfer: warm faded grade, hard sun, period cars and people; brick and terracotta from image 2; massing recognisable with a little drift.", "latencyMs": 61966, "method": "original blind review" }, { "key": "flux-2-pro", "name": "Flux 2 Pro", "credits": 100, "score": 5, "dims": { "design": 8, "style": 5, "prompt": 6, "aesthetic": 6 }, "note": "Design and view held, brick right, but cool hard-contrast grade with little Kodachrome character.", "latencyMs": 23077, "method": "original blind review" } ] }, { "id": "R4-render-json-interior", "capability": "render_json_no_ref", "prompt": "{\n \"task\": \"Render this line drawing as a photorealistic interior photograph, keeping the exact camera, framing and aspect ratio, with every object, wall, window and cupboard where the drawing has it\",\n \"materials\": {\n \"floor\": \"polished concrete, light grey, low sheen\",\n \"walls_and_ceiling\": \"white plaster, matte\",\n \"joinery\": \"pale oak veneer cupboards with shadow-gap pulls\",\n \"window_frames\": \"slim matte black steel\",\n \"fireplace\": \"matte black steel, suspended from the ceiling\",\n \"ottomans\": \"charcoal wool boucle\",\n \"lounge_chairs\": \"black leather seat and back on a light oak frame\",\n \"table_and_stool\": \"solid pale oak\"\n },\n \"lighting\": {\n \"source\": \"soft overcast daylight through the glazing on both sides\",\n \"time_of_day\": \"mid morning\",\n \"artificial_lights\": \"off\",\n \"shadows\": \"soft, low contrast\"\n },\n \"photography\": {\n \"camera\": \"full-frame, 24mm shift lens, verticals perfectly straight\",\n \"exposure\": \"bright and airy with gently lifted shadows\",\n \"depth_of_field\": \"deep, everything in focus\",\n \"style\": \"editorial architectural interior photograph, Dezeen or The Local Project\"\n },\n \"colour\": {\n \"palette\": \"white, pale oak, light grey concrete, black accents\",\n \"grade\": \"neutral, natural white balance\"\n },\n \"outside\": \"green garden foliage seen through the glazing, slightly soft\"\n}", "criteria": "Positional fidelity to the drawing first (read the blend row), then how many of the JSON fields were honoured - materials by element, lighting, camera and grade - then whether it reads as an editorial photograph. NO reference image: style adherence here means adherence to the written photographic style, not to a picture. Treat any object, opening or wall moved from where the drawing has it as a fault even when the image is beautiful.", "inputs": [ "linework_interior@4x3" ], "overlay": true, "cells": [ { "key": "runchat-image", "name": "Runchat Image", "credits": 4, "score": 5, "dims": { "design": 7, "style": 6, "prompt": 7, "aesthetic": 5 }, "note": "Positions roughly held but the left block is white upholstery, ottomans mixed grey, and the whole image reads as a flat render.", "latencyMs": 6146, "method": "original blind review" }, { "key": "muse-image", "name": "Muse Image", "credits": 20, "score": 9, "dims": { "design": 9, "style": 8, "prompt": 8, "aesthetic": 9 }, "note": "Camera held, clerestory and paired oak cupboards as drawn, charcoal ottomans, black-and-oak chairs, soft daylight; the most editorial result. Left block only partly oak.", "latencyMs": 21755, "method": "original blind review" }, { "key": "gpt-image-2-5-sunburst-low", "name": "GPT Image 2.5 (low)", "credits": 40, "score": 8, "dims": { "design": 7, "style": 9, "prompt": 8, "aesthetic": 9 }, "note": "Correct concrete, oak, charcoal upholstery and black steel with convincing soft daylight. Slightly changes furniture placement, fireplace position and the framing compared with the line drawing.", "latencyMs": 19280, "method": "non-blind follow-up" }, { "key": "nano-banana-1", "name": "Nano Banana 1", "credits": 80, "score": 8, "dims": { "design": 8, "style": 7, "prompt": 9, "aesthetic": 7 }, "note": "Most literal to the JSON: oak block, oak plinths under the ottomans, clerestory, paired cupboards; a little flatter and softer than the best.", "latencyMs": 9717, "method": "original blind review" }, { "key": "nano-banana-pro-2k", "name": "Nano Banana Pro (2K)", "credits": 300, "score": 7, "dims": { "design": 8, "style": 8, "prompt": 7, "aesthetic": 9 }, "note": "Excellent realism and lighting, clerestory and cupboards right, but the left block became charcoal boucle instead of oak.", "latencyMs": 30220, "method": "original blind review" }, { "key": "flux-2-pro", "name": "Flux 2 Pro", "credits": 100, "score": 7, "dims": { "design": 8, "style": 8, "prompt": 6, "aesthetic": 9 }, "note": "Most beautiful image of the six, but the desk group turned black leather and the chairs lost their oak frames; a curtain was added.", "latencyMs": 18758, "method": "original blind review" } ] }, { "id": "S1-site-short", "capability": "site_placement", "prompt": "Place the building in image 1 inside the red outline in image 2 and remove the outline.", "criteria": "Placement inside the outline first, then scale against the neighbours, then light and shadow consistency, then how well the design of image 1 survives (balconied long facades, hipped roof, stair core), then preservation of the rest of the photo. SHORT prompt: measures what the model does unprompted about scale, light and material.", "inputs": [ "revit_midrise", "site_aerial_lot_marked@4x3" ], "overlay": true, "cells": [ { "key": "runchat-image", "name": "Runchat Image", "credits": 4, "score": 6, "dims": { "design": 7, "style": 5, "prompt": 7, "aesthetic": 6 }, "note": "Placed inside the lot on a plinth, outline removed, a little small; pasted look, no shadow.", "latencyMs": 7992, "method": "original blind review" }, { "key": "muse-image", "name": "Muse Image", "credits": 20, "score": 8, "dims": { "design": 8, "style": 6, "prompt": 8, "aesthetic": 7 }, "note": "Inside the outline at the right scale, outline gone, design intact, a hint of ground shadow; still reads as a render pasted on a photo.", "latencyMs": 21214, "method": "original blind review" }, { "key": "gpt-image-2-5-sunburst-low", "name": "GPT Image 2.5 (low)", "credits": 40, "score": 6, "dims": { "design": 8, "style": 5, "prompt": 6, "aesthetic": 5 }, "note": "Removes the outline and places the recognisable building in the lot while keeping the surrounding photograph. Turquoise roof and diagram-like facade remain conspicuously rendered; contact and cast-shadow integration are weak.", "latencyMs": 22656, "method": "non-blind follow-up" }, { "key": "nano-banana-1", "name": "Nano Banana 1", "credits": 80, "score": 4, "dims": { "design": 5, "style": 5, "prompt": 5, "aesthetic": 6 }, "note": "Drifted right onto the edge of the circular tank, outside the outlined lot.", "latencyMs": 11152, "method": "original blind review" }, { "key": "nano-banana-pro-2k", "name": "Nano Banana Pro (2K)", "credits": 300, "score": 5, "dims": { "design": 7, "style": 4, "prompt": 6, "aesthetic": 5 }, "note": "Inside the outline but pasted with its own perspective, so the camera does not match the aerial.", "latencyMs": 27814, "method": "original blind review" }, { "key": "flux-2-pro", "name": "Flux 2 Pro", "credits": 100, "score": 3, "dims": { "design": 4, "style": 4, "prompt": 4, "aesthetic": 5 }, "note": "Right lot, outline removed, but toy scale: a fraction of the neighbouring blocks.", "latencyMs": 34380, "method": "original blind review" } ] }, { "id": "S2-site-detailed", "capability": "site_placement", "prompt": "Place the building in image 1 inside the red outline in image 2, filling the outline with its long balconied facade facing the parking row at the bottom of the frame, and remove the outline. Scale: the same height as the neighbouring five-storey apartment blocks. Light: the same sun direction, shadow length and shadow softness as the existing buildings, with the shadow falling across the lot the way theirs do. Material: pale render and warm terracotta roof tiles in the same slightly dusty tones as the neighbouring buildings so it reads as part of the same street. Keep the viewpoint of image 2 and everything outside the outline exactly as photographed.", "criteria": "Same as S1 with the DETAILED prompt: placement inside the outline, orientation as instructed, scale, light and shadow, material tone, then preservation. Compare the pair to see what the instructions buy.", "inputs": [ "revit_midrise", "site_aerial_lot_marked@4x3" ], "overlay": true, "cells": [ { "key": "runchat-image", "name": "Runchat Image", "credits": 4, "score": 5, "dims": { "design": 6, "style": 7, "prompt": 5, "aesthetic": 7 }, "note": "Materials and scale right but the outline was left in place and the shadow is weak.", "latencyMs": 7560, "method": "original blind review" }, { "key": "muse-image", "name": "Muse Image", "credits": 20, "score": 7, "dims": { "design": 6, "style": 8, "prompt": 7, "aesthetic": 8 }, "note": "Best integration: terracotta roof, pale render, matching sun and shadow, outline removed; but the block was redesigned as an L-shaped hipped building and grew past the neighbours.", "latencyMs": 34826, "method": "original blind review" }, { "key": "gpt-image-2-5-sunburst-low", "name": "GPT Image 2.5 (low)", "credits": 40, "score": 5, "dims": { "design": 5, "style": 7, "prompt": 4, "aesthetic": 6 }, "note": "Adopts dusty pale render and a terracotta roof and removes the outline. Redesigns the roof and retains a tall block rather than matching the five-storey neighbours; the long balconied facade does not face the parking row as requested.", "latencyMs": 22618, "method": "non-blind follow-up" }, { "key": "nano-banana-1", "name": "Nano Banana 1", "credits": 80, "score": 4, "dims": { "design": 7, "style": 4, "prompt": 4, "aesthetic": 5 }, "note": "Render pasted unchanged: cyan roof, no shadow, no material change.", "latencyMs": 11169, "method": "original blind review" }, { "key": "nano-banana-pro-2k", "name": "Nano Banana Pro (2K)", "credits": 300, "score": 6, "dims": { "design": 5, "style": 8, "prompt": 6, "aesthetic": 8 }, "note": "Terracotta and render with convincing shadows, but the red outline is still on the image and the design became generic.", "latencyMs": 36404, "method": "original blind review" }, { "key": "flux-2-pro", "name": "Flux 2 Pro", "credits": 100, "score": 6, "dims": { "design": 7, "style": 5, "prompt": 7, "aesthetic": 6 }, "note": "Design kept, outline removed, a real cast shadow; roof still cyan and the whole thing reads as a render.", "latencyMs": 18802, "method": "original blind review" } ] }, { "id": "O1-insert-short", "capability": "object_insertion_lighting", "prompt": "Add the sofa from image 2 to the room in image 1, standing in the sunlight in front of the window.", "criteria": "Lighting first: does the sofa receive the same hard low sun as the room, with a shadow of the right direction, length and sharpness and a contact shadow. Then scale and perspective on the boards, then fidelity to the sofa design, then preservation of the room. SHORT prompt: measures what the model works out from the photo alone.", "inputs": [ "room_sunlit_empty@4x3", "product_togo_lounge" ], "overlay": true, "cells": [ { "key": "runchat-image", "name": "Runchat Image", "credits": 4, "score": 7, "dims": { "design": 6, "style": 8, "prompt": 7, "aesthetic": 8 }, "note": "Good light on the window side and a shadow to the right, but the sofa became a single-seat Togo.", "latencyMs": 8369, "method": "original blind review" }, { "key": "muse-image", "name": "Muse Image", "credits": 20, "score": 8, "dims": { "design": 8, "style": 8, "prompt": 8, "aesthetic": 8 }, "note": "Two-seater kept, sits in the light patch with sun on the leather and a soft cast shadow to the right; convincing.", "latencyMs": 17793, "method": "original blind review" }, { "key": "gpt-image-2-5-sunburst-low", "name": "GPT Image 2.5 (low)", "credits": 40, "score": 9, "dims": { "design": 9, "style": 9, "prompt": 9, "aesthetic": 9 }, "note": "Preserves the room and inserts a convincing tan two-seat Togo in the sun patch. Leather channels, three-quarter form, bright window-side light and floor contact are strong.", "latencyMs": 23259, "method": "non-blind follow-up" }, { "key": "nano-banana-1", "name": "Nano Banana 1", "credits": 80, "score": 7, "dims": { "design": 6, "style": 7, "prompt": 7, "aesthetic": 7 }, "note": "Single-seater with plausible highlights and a shadow behind it; room preserved.", "latencyMs": 12186, "method": "original blind review" }, { "key": "nano-banana-pro-2k", "name": "Nano Banana Pro (2K)", "credits": 300, "score": 8, "dims": { "design": 9, "style": 7, "prompt": 8, "aesthetic": 8 }, "note": "Two-seater kept and well placed with a contact shadow; lighting a touch even, the cast shadow faint.", "latencyMs": 29152, "method": "original blind review" }, { "key": "flux-2-pro", "name": "Flux 2 Pro", "credits": 100, "score": 4, "dims": { "design": 6, "style": 4, "prompt": 5, "aesthetic": 5 }, "note": "Single-seater facing the camera with flat light and no cast shadow; reads as pasted.", "latencyMs": 21443, "method": "original blind review" } ] }, { "id": "O2-insert-multi", "capability": "object_insertion_lighting", "prompt": "Furnish the room in image 1 with the pieces in images 2 to 4. Place the sofa from image 2 on the floorboards inside the patch of direct sunlight in front of the arched window, turned three-quarters toward the camera, at true scale for the room. Lay the rug from image 3 flat on the floor under and in front of the sofa, its pattern foreshortened correctly along the boards. Put the brass side table and the striped table lamp from image 4 beside one end of the sofa. Sunlight enters low through the window from the left, so every piece is brightly lit on its window side and in shade on the far side, each casting a hard-edged shadow across the boards at the same angle and sharpness as the light patch from the window, with contact shadows where they meet the floor and the rug. Keep the design, materials and colours of each piece exactly as in its image, and keep the room, camera and framing exactly as photographed.", "criteria": "Completeness and placement first (all three pieces, sofa in the light, rug under it, table beside it), then lighting consistency across all of them (one sun, matching shadows), then fidelity of each piece to its image, then scale and perspective, then preservation of the room. Compare with O1 to see what three objects at once cost.", "inputs": [ "room_sunlit_empty@4x3", "product_togo_lounge", "product_rug", "product_cocoflip" ], "overlay": true, "cells": [ { "key": "runchat-image", "name": "Runchat Image", "credits": 4, "score": 8, "dims": { "design": 9, "style": 8, "prompt": 8, "aesthetic": 7 }, "note": "All three present and lit from the window with the rug pattern kept; slightly softer.", "latencyMs": 13406, "method": "original blind review" }, { "key": "muse-image", "name": "Muse Image", "credits": 20, "score": 8, "dims": { "design": 9, "style": 8, "prompt": 8, "aesthetic": 8 }, "note": "All three placed and lit, rug pattern faithful; the sofa is a touch large and its own shadow is weak.", "latencyMs": 32008, "method": "original blind review" }, { "key": "gpt-image-2-5-sunburst-low", "name": "GPT Image 2.5 (low)", "credits": 40, "score": 9, "dims": { "design": 9, "style": 9, "prompt": 9, "aesthetic": 9 }, "note": "Keeps the sofa, patterned rug, brass table and striped lamp recognisable at plausible scale. Strong window-side light and coherent rug perspective; room and camera remain close to the input.", "latencyMs": 25209, "method": "non-blind follow-up" }, { "key": "nano-banana-1", "name": "Nano Banana 1", "credits": 80, "score": 3, "dims": { "design": 6, "style": 7, "prompt": 4, "aesthetic": 7 }, "note": "Replaced the room with a different one; the pieces are fine but the photograph is gone.", "latencyMs": 13931, "method": "original blind review" }, { "key": "nano-banana-pro-2k", "name": "Nano Banana Pro (2K)", "credits": 300, "score": 9, "dims": { "design": 9, "style": 9, "prompt": 9, "aesthetic": 9 }, "note": "All three pieces true to their images, sofa in the sun with the window grid falling across the rug, table and lamp beside it; camera held.", "latencyMs": 34406, "method": "original blind review" }, { "key": "flux-2-pro", "name": "Flux 2 Pro", "credits": 100, "score": 8, "dims": { "design": 9, "style": 8, "prompt": 8, "aesthetic": 7 }, "note": "All three placed, lamp cord and all, rug faithful; lighting a little even.", "latencyMs": 30432, "method": "original blind review" } ] }, { "id": "E1-edit-short", "capability": "local_edit", "prompt": "Change the central joinery box to dark stained oak.", "criteria": "Region discipline first: only the freestanding box changes, the same-finish kitchen cupboards behind it and every other surface stay - read the blend row; then completeness of the change on all visible faces; then material believability and lighting. SHORT prompt: measures whether the model can tell the named element from its same-material neighbour unprompted.", "inputs": [ "divisare_interior2@4x3" ], "overlay": true, "cells": [ { "key": "runchat-image", "name": "Runchat Image", "credits": 4, "score": 4, "dims": { "design": 4, "style": 6, "prompt": 4, "aesthetic": 6 }, "note": "Box recut into framed shaker panels and the kitchen darkened with it.", "latencyMs": 5634, "method": "original blind review" }, { "key": "muse-image", "name": "Muse Image", "credits": 20, "score": 4, "dims": { "design": 4, "style": 7, "prompt": 4, "aesthetic": 7 }, "note": "Panelled doors on the box and the kitchen both turned dark oak.", "latencyMs": 14929, "method": "original blind review" }, { "key": "gpt-image-2-5-sunburst-low", "name": "GPT Image 2.5 (low)", "credits": 40, "score": 5, "dims": { "design": 8, "style": 8, "prompt": 4, "aesthetic": 8 }, "note": "Darkens the central box convincingly but also changes the kitchen cupboards behind it to dark timber. This fails the central requirement to change only the named joinery.", "latencyMs": 18073, "method": "non-blind follow-up" }, { "key": "nano-banana-1", "name": "Nano Banana 1", "credits": 80, "score": 5, "dims": { "design": 5, "style": 6, "prompt": 5, "aesthetic": 6 }, "note": "Box dark oak with the drawn divisions, kitchen also darkened; slightly soft.", "latencyMs": 8351, "method": "original blind review" }, { "key": "nano-banana-pro-2k", "name": "Nano Banana Pro (2K)", "credits": 300, "score": 8, "dims": { "design": 8, "style": 9, "prompt": 8, "aesthetic": 9 }, "note": "Only the box changed and the kitchen run stayed pale grey; superb dark oak, though an open niche was added to the box.", "latencyMs": 34352, "method": "original blind review" }, { "key": "flux-2-pro", "name": "Flux 2 Pro", "credits": 100, "score": 7, "dims": { "design": 7, "style": 7, "prompt": 8, "aesthetic": 7 }, "note": "Kitchen kept pale grey, box dark oak; door layout and handles were redesigned.", "latencyMs": 14196, "method": "original blind review" } ] }, { "id": "E2-edit-multi-material", "capability": "local_edit", "prompt": "Apply these materials to the room in image 1. The freestanding joinery box in the centre: the horizontal larch slats of image 2. The kitchen cupboards and the wall behind them on the right: the patterned red and white tiles of image 3. The floor: the pale limestone flags of image 4. The two doors on the left wall: the charred black timber of image 5. Walls, ceiling and the cast-iron column stay white. Keep the camera and framing of image 1.", "criteria": "Assignment discipline first: each reference lands on its named element and nowhere else - read the blend row for anything that moved; then how faithfully each material is reproduced (pattern, scale, colour); then lighting consistency on the new surfaces; then overall quality. Compare with E1 to see how much four simultaneous assignments cost.", "inputs": [ "divisare_interior2@4x3", "divisare_timber2", "divisare_tiles", "divisare_stone", "divisare_timber3" ], "overlay": true, "cells": [ { "key": "runchat-image", "name": "Runchat Image", "credits": 4, "score": 6, "dims": { "design": 7, "style": 7, "prompt": 7, "aesthetic": 6 }, "note": "Assignments mostly right; the kitchen base cupboards came out oak and the image is soft.", "latencyMs": 9894, "method": "original blind review" }, { "key": "muse-image", "name": "Muse Image", "credits": 20, "score": 9, "dims": { "design": 9, "style": 9, "prompt": 9, "aesthetic": 8 }, "note": "Every assignment on its element and nowhere else: larch box, tiled kitchen recess, limestone floor, charred doors, white walls, camera held.", "latencyMs": 23878, "method": "original blind review" }, { "key": "gpt-image-2-5-sunburst-low", "name": "GPT Image 2.5 (low)", "credits": 40, "score": 9, "dims": { "design": 9, "style": 8, "prompt": 9, "aesthetic": 9 }, "note": "Assigns horizontal slats, patterned tiles, limestone flooring and black doors to the correct elements, with white walls and column retained. Tile red is subdued, but the camera and daylight are well preserved.", "latencyMs": 24486, "method": "non-blind follow-up" }, { "key": "nano-banana-1", "name": "Nano Banana 1", "credits": 80, "score": 3, "dims": { "design": 3, "style": 6, "prompt": 5, "aesthetic": 6 }, "note": "Camera moved and tiles cover several walls; the box became a freestanding slatted column.", "latencyMs": 12834, "method": "original blind review" }, { "key": "nano-banana-pro-2k", "name": "Nano Banana Pro (2K)", "credits": 300, "score": 8, "dims": { "design": 7, "style": 9, "prompt": 8, "aesthetic": 9 }, "note": "Most beautiful materials, but the tiles spread onto the wall between the balcony doors.", "latencyMs": 44484, "method": "original blind review" }, { "key": "flux-2-pro", "name": "Flux 2 Pro", "credits": 100, "score": 4, "dims": { "design": 4, "style": 7, "prompt": 6, "aesthetic": 6 }, "note": "Materials applied but the room was re-laid: doors moved, windows merged, tiles on extra walls.", "latencyMs": 22245, "method": "original blind review" } ] }, { "id": "V1-view-sheet", "capability": "novel_views", "prompt": "Produce a 2x2 sheet of orthographic drawings of this house in the same clean white model style as the input, drawings only. Top-left: the front elevation, the end with the tall fully glazed gable. Top-right: the right side elevation, showing the long side of the gabled volume and the lower flat-roofed glazed wing extending from it. Bottom-left: the rear elevation. Bottom-right: the roof plan seen from directly above, showing the pitched roof and the flat roof of the wing. Every view strictly orthographic with parallel edges, all four at the same scale and consistent with the geometry visible in the input.", "criteria": "Cross-view consistency first: do the four drawings describe one building that matches the input; then correctness of each view (orthographic, right side, right roof form); then layout as instructed; then line quality.", "inputs": [ "shaded_barn" ], "overlay": false, "cells": [ { "key": "runchat-image", "name": "Runchat Image", "credits": 4, "score": 5, "dims": { "design": 5, "style": 7, "prompt": 5, "aesthetic": 6 }, "note": "Front and rear fine, but the side view is a perspective and the roof plan is hipped with dimension text.", "latencyMs": 4839, "method": "original blind review" }, { "key": "muse-image", "name": "Muse Image", "credits": 20, "score": 8, "dims": { "design": 8, "style": 8, "prompt": 8, "aesthetic": 8 }, "note": "Four true orthographic views in the stated positions, mutually consistent, gable ridge and flat wing right in the roof plan; only breach is the printed labels. (Second on review.)", "latencyMs": 20450, "method": "original blind review" }, { "key": "gpt-image-2-5-sunburst-low", "name": "GPT Image 2.5 (low)", "credits": 40, "score": 8, "dims": { "design": 8, "style": 9, "prompt": 8, "aesthetic": 8 }, "note": "Produces the four requested orthographic views in the correct positions with a coherent roof plan and white model finish. Some side and rear openings are simplified or inferred, so this remains a concept drawing rather than verified geometry.", "latencyMs": 23021, "method": "non-blind follow-up" }, { "key": "nano-banana-1", "name": "Nano Banana 1", "credits": 80, "score": 3, "dims": { "design": 3, "style": 4, "prompt": 3, "aesthetic": 4 }, "note": "Faint, inconsistent drawings of a different boxy house.", "latencyMs": 9259, "method": "original blind review" }, { "key": "nano-banana-pro-2k", "name": "Nano Banana Pro (2K)", "credits": 300, "score": 9, "dims": { "design": 9, "style": 9, "prompt": 9, "aesthetic": 9 }, "note": "Detailed and consistent, standing-seam roof and stilted wing carried through all four views; the plinth is drawn in slight perspective. (Moved to first on review: the most accurate and complete sheet.)", "latencyMs": 33173, "method": "original blind review" }, { "key": "flux-2-pro", "name": "Flux 2 Pro", "credits": 100, "score": 4, "dims": { "design": 4, "style": 6, "prompt": 4, "aesthetic": 5 }, "note": "Perspective creeping into the elevations, a hipped roof plan and garbled label text.", "latencyMs": 15647, "method": "original blind review" } ] }, { "id": "C1-merge-short", "capability": "concept_merge", "prompt": "Create a new object that is a formal hybrid of these two, halfway between them in shape. Image 1 is a thin continuous surface that folds back on itself with no inside or outside, opening into arches and holes. Image 2 is a closed, squat ovoid shell with one large round opening. Matte white, on a white background.", "criteria": "Morph first: a shape that sits between the two topologies beats either parent, and beats a parent wearing the other one as texture; then whether both the folding-surface logic and the closed-ovoid massing are legible in the geometry; then object coherence and photographic quality. Purely formal, material neutral. SHORT prompt.", "inputs": [ "ref_eva_hild", "ref_un_studio" ], "overlay": false, "cells": [ { "key": "runchat-image", "name": "Runchat Image", "credits": 4, "score": 6, "dims": { "design": 6, "style": 7, "prompt": 6, "aesthetic": 7 }, "note": "Looping folded surfaces with two ring openings; the fold logic is there but the closed ovoid massing is weak.", "latencyMs": 6575, "method": "original blind review" }, { "key": "muse-image", "name": "Muse Image", "credits": 20, "score": 9, "dims": { "design": 9, "style": 9, "prompt": 8, "aesthetic": 9 }, "note": "Ovoid mass carrying folds, arches and holes across it; a true morph, set on a grey ground rather than white.", "latencyMs": 20059, "method": "original blind review" }, { "key": "gpt-image-2-5-sunburst-low", "name": "GPT Image 2.5 (low)", "credits": 40, "score": 9, "dims": { "design": 9, "style": 9, "prompt": 9, "aesthetic": 9 }, "note": "A convincing new squat ovoid with a large rolled opening and continuous curling surfaces and holes. Integrates both source forms without retaining the pod's literal skin.", "latencyMs": 14744, "method": "non-blind follow-up" }, { "key": "nano-banana-1", "name": "Nano Banana 1", "credits": 80, "score": 3, "dims": { "design": 3, "style": 5, "prompt": 3, "aesthetic": 7 }, "note": "The Hild sculpture reproduced with no ovoid massing.", "latencyMs": 12001, "method": "original blind review" }, { "key": "nano-banana-pro-2k", "name": "Nano Banana Pro (2K)", "credits": 300, "score": 9, "dims": { "design": 9, "style": 9, "prompt": 9, "aesthetic": 9 }, "note": "A closed ovoid with one large opening whose shell peels into thin folding surfaces on one side; neither parent on its own; matte white on white.", "latencyMs": 27530, "method": "original blind review" }, { "key": "flux-2-pro", "name": "Flux 2 Pro", "credits": 100, "score": 5, "dims": { "design": 5, "style": 6, "prompt": 6, "aesthetic": 6 }, "note": "A squat shell with a round opening and perforations; reads as a product casing, folds reduced to one flap.", "latencyMs": 18753, "method": "original blind review" } ] }, { "id": "C2-merge-detailed", "capability": "concept_merge", "prompt": "Design a new object whose geometry morphs between image 1 and image 2, halfway between them, so that neither original is recognisable on its own. Take from image 1 its topology: one thin continuous surface that folds back on itself, turns inside out, and opens into arches and holes, with no clear inside or outside. Take from image 2 its massing: a squat, grounded, closed ovoid shell with a single large round opening and a thick rounded rim. The result should read as an ovoid that has begun to peel open into folding surfaces, or as a folded surface that has closed itself into a near-egg, with the rim of image 2 becoming one of the folds of image 1. Render it in plain matte white plaster so that only the form counts: a studio photograph of the object alone on a white background in soft even light.", "criteria": "Same as C1 with the DETAILED prompt: the topology of image 1, the massing of image 2 and the morph mechanism are spelled out, so score how far the geometry actually sits between the two. Purely formal, material neutral. Compare the pair.", "inputs": [ "ref_eva_hild", "ref_un_studio" ], "overlay": false, "cells": [ { "key": "runchat-image", "name": "Runchat Image", "credits": 4, "score": 5, "dims": { "design": 5, "style": 5, "prompt": 5, "aesthetic": 7 }, "note": "Sphere with one hole and a fold line; the ovoid with almost no folding topology.", "latencyMs": 6301, "method": "original blind review" }, { "key": "muse-image", "name": "Muse Image", "credits": 20, "score": 8, "dims": { "design": 8, "style": 8, "prompt": 8, "aesthetic": 8 }, "note": "Rim and bowl of image 2 sitting on a folded, holed mass; strong but the two halves are still a little separable.", "latencyMs": 29288, "method": "original blind review" }, { "key": "gpt-image-2-5-sunburst-low", "name": "GPT Image 2.5 (low)", "credits": 40, "score": 9, "dims": { "design": 9, "style": 9, "prompt": 9, "aesthetic": 9 }, "note": "Combines a grounded near-egg mass with a rolled opening and thin surfaces peeling into arches and holes. The two sources are integrated into one coherent matte-white object.", "latencyMs": 23344, "method": "non-blind follow-up" }, { "key": "nano-banana-1", "name": "Nano Banana 1", "credits": 80, "score": 4, "dims": { "design": 4, "style": 6, "prompt": 4, "aesthetic": 7 }, "note": "The Hild sculpture with a torus ring added; a collage rather than a morph.", "latencyMs": 11764, "method": "original blind review" }, { "key": "nano-banana-pro-2k", "name": "Nano Banana Pro (2K)", "credits": 300, "score": 9, "dims": { "design": 9, "style": 9, "prompt": 9, "aesthetic": 9 }, "note": "Squat ovoid with the thick rounded rim of image 2 that becomes the first of a series of folds with holes; exactly the stated morph.", "latencyMs": 27779, "method": "original blind review" }, { "key": "flux-2-pro", "name": "Flux 2 Pro", "credits": 100, "score": 5, "dims": { "design": 5, "style": 6, "prompt": 5, "aesthetic": 7 }, "note": "Ovoid massing with rim and opening, but the folds were replaced by a perforated skin: texture standing in for topology.", "latencyMs": 17304, "method": "original blind review" } ] }, { "id": "M1-material-short", "capability": "material_transfer", "prompt": "Build the sculpture in image 1 as a full-scale woven rattan pavilion standing inside the shed in image 2, with areas of solid weave and areas of open bundled cane.", "criteria": "Scale and setting first: a pavilion in the shed, not a tabletop object; then fidelity to the form of image 1; then whether both construction modes are present and placed sensibly (solid weave on the bodies, open bundles at the edges and openings); then realism. SHORT prompt.", "inputs": [ "sculpture_merged", "ref_rattan" ], "overlay": false, "cells": [ { "key": "runchat-image", "name": "Runchat Image", "credits": 4, "score": 3, "dims": { "design": 2, "style": 6, "prompt": 3, "aesthetic": 6 }, "note": "Reproduced the reference pavilion from image 2 with a person; the form of image 1 is absent.", "latencyMs": 6530, "method": "original blind review" }, { "key": "muse-image", "name": "Muse Image", "credits": 20, "score": 9, "dims": { "design": 9, "style": 9, "prompt": 9, "aesthetic": 9 }, "note": "The sculpture at pavilion scale in the shed: solid woven bodies, perforated open weave on the folds, bound rims; form and holes exact.", "latencyMs": 42072, "method": "original blind review" }, { "key": "gpt-image-2-5-sunburst-low", "name": "GPT Image 2.5 (low)", "credits": 40, "score": 9, "dims": { "design": 9, "style": 9, "prompt": 9, "aesthetic": 9 }, "note": "Rebuilds the sculpture's folds and circular opening at pavilion scale, with solid woven panels and open cane transitions. The workshop setting and person make the scale legible.", "latencyMs": 16620, "method": "non-blind follow-up" }, { "key": "nano-banana-1", "name": "Nano Banana 1", "credits": 80, "score": 5, "dims": { "design": 4, "style": 7, "prompt": 5, "aesthetic": 7 }, "note": "A pavilion in the shed with solid and open weave, but the form is a cluster of pods rather than the sculpture.", "latencyMs": 11963, "method": "original blind review" }, { "key": "nano-banana-pro-2k", "name": "Nano Banana Pro (2K)", "credits": 300, "score": 9, "dims": { "design": 9, "style": 9, "prompt": 8, "aesthetic": 9 }, "note": "Same form at even larger scale in bundled coiled cane with open loops in the ring; superb realism.", "latencyMs": 31382, "method": "original blind review" }, { "key": "flux-2-pro", "name": "Flux 2 Pro", "credits": 100, "score": 2, "dims": { "design": 2, "style": 3, "prompt": 2, "aesthetic": 5 }, "note": "Pasted the white sculpture into a copy of the reference pavilion; a collage.", "latencyMs": 31620, "method": "original blind review" } ] }, { "id": "M2-material-detailed", "capability": "material_transfer", "prompt": "Build the sculpture in image 1 as a full-scale woven rattan pavilion, large enough to walk into, standing on the concrete floor inside the shed in image 2 and lit by the same daylight from its openings. Keep the exact form of image 1: the same folds, ridges, holes and proportions, only scaled up to about four metres tall. Construct it from rattan the way image 2 is made: the broad bodies of the form as solid, tightly woven surfaces; the folded edges, ridges and the rims of the holes as thick bundles of cane bound together; and the thin transitional areas between them left as open bundled cane with no solid surface, so light passes through and the structure shows. Natural rattan colour. A photograph taken inside the shed with a person standing beside the pavilion for scale.", "criteria": "Same as M1 with the DETAILED prompt: scale, setting, form fidelity and the assignment of solid weave, bound bundles and open bundles are all specified, so score execution of each. Compare the pair.", "inputs": [ "sculpture_merged", "ref_rattan" ], "overlay": false, "cells": [ { "key": "runchat-image", "name": "Runchat Image", "credits": 4, "score": 3, "dims": { "design": 2, "style": 6, "prompt": 3, "aesthetic": 6 }, "note": "Copied the reference pavilion with a person; form of image 1 absent.", "latencyMs": 9311, "method": "original blind review" }, { "key": "muse-image", "name": "Muse Image", "credits": 20, "score": 9, "dims": { "design": 9, "style": 9, "prompt": 9, "aesthetic": 9 }, "note": "Form exact, intricate mix of solid weave, open bundles and bound rims, person for scale; equally complete.", "latencyMs": 33829, "method": "original blind review" }, { "key": "gpt-image-2-5-sunburst-low", "name": "GPT Image 2.5 (low)", "credits": 40, "score": 9, "dims": { "design": 9, "style": 9, "prompt": 9, "aesthetic": 9 }, "note": "Preserves the major folds and ring, with woven bodies, bundled rims and open cane in the transitions. Person, concrete floor and shed lighting establish a convincing full-scale installation.", "latencyMs": 16994, "method": "non-blind follow-up" }, { "key": "nano-banana-1", "name": "Nano Banana 1", "credits": 80, "score": 7, "dims": { "design": 8, "style": 8, "prompt": 7, "aesthetic": 8 }, "note": "Form kept with solid bodies and an open window, rims understated; realistic.", "latencyMs": 11951, "method": "original blind review" }, { "key": "nano-banana-pro-2k", "name": "Nano Banana Pro (2K)", "credits": 300, "score": 9, "dims": { "design": 9, "style": 9, "prompt": 9, "aesthetic": 9 }, "note": "Exact form at four metres with a person, solid woven bodies, bound bundle rims and open lattice between; every instruction delivered.", "latencyMs": 30432, "method": "original blind review" }, { "key": "flux-2-pro", "name": "Flux 2 Pro", "credits": 100, "score": 3, "dims": { "design": 2, "style": 6, "prompt": 3, "aesthetic": 6 }, "note": "Copied the reference pavilion with a person; form of image 1 absent.", "latencyMs": 19940, "method": "original blind review" } ] }, { "id": "G1-icons-six", "capability": "graphics_icons", "prompt": "Study the brand kit in image 1: the arch logo drawn as nested thin parallel lines with rounded corners, and the coral and periwinkle swirl motif of flowing parallel ribbons. Design six app icons for this brand, house, key, map pin, calendar, phone and search, each constructed the same way as the logo, from nested parallel lines with rounded arches, with one flowing swirl stroke worked into each, in the kit palette on the deep purple.", "criteria": "Interpretation of the kit first: do the icons share the logo construction (nested parallel lines, rounded arches) and carry the swirl motif, rather than being generic line icons placed on a purple tile; then recognisability of all six; then consistency. SHORT prompt.", "inputs": [ "brand_kit" ], "overlay": false, "cells": [ { "key": "runchat-image", "name": "Runchat Image", "credits": 4, "score": 7, "dims": { "design": 9, "style": 8, "prompt": 7, "aesthetic": 7 }, "note": "Nested-line construction and swirls done well, but purple on a light grey ground instead of the kit purple.", "latencyMs": 5430, "method": "original blind review" }, { "key": "muse-image", "name": "Muse Image", "credits": 20, "score": 8, "dims": { "design": 9, "style": 9, "prompt": 7, "aesthetic": 9 }, "note": "Nested white lines with coral and periwinkle ribbons, the closest reading of the motif; printed labels under each icon breach icons-only.", "latencyMs": 20895, "method": "original blind review" }, { "key": "gpt-image-2-5-sunburst-low", "name": "GPT Image 2.5 (low)", "credits": 40, "score": 8, "dims": { "design": 9, "style": 7, "prompt": 8, "aesthetic": 8 }, "note": "All six icons are recognisable with parallel outlines and coral curls on purple. Strokes are heavier than the brand's thin-line construction, and rounded-square tiles add an unrequested container.", "latencyMs": 18744, "method": "non-blind follow-up" }, { "key": "nano-banana-1", "name": "Nano Banana 1", "credits": 80, "score": 4, "dims": { "design": 8, "style": 4, "prompt": 4, "aesthetic": 6 }, "note": "Generic white line icons on tiles with the swirl motif as wallpaper; the icons themselves ignore the logo construction.", "latencyMs": 10122, "method": "original blind review" }, { "key": "nano-banana-pro-2k", "name": "Nano Banana Pro (2K)", "credits": 300, "score": 9, "dims": { "design": 9, "style": 9, "prompt": 9, "aesthetic": 9 }, "note": "Every icon built from nested coral lines with a periwinkle swirl worked through it on the deep purple; reads as the same family as the logo.", "latencyMs": 35719, "method": "original blind review" }, { "key": "flux-2-pro", "name": "Flux 2 Pro", "credits": 100, "score": 6, "dims": { "design": 8, "style": 7, "prompt": 7, "aesthetic": 6 }, "note": "Double-line icons with swirl trails on purple tiles; the nesting is shallow and the swirls sit beside the icons.", "latencyMs": 23368, "method": "original blind review" } ] }, { "id": "G2-icons-twentyfour", "capability": "graphics_icons", "prompt": "Study the brand kit in image 1 before drawing anything: the logo is an arch drawn as three nested thin parallel lines with rounded corners, and the motif is a set of flowing parallel ribbons in coral and periwinkle that curl across the ground. Design a set of twenty-four app icons for this brand in a 6x4 grid with even spacing, in this order left to right, top to bottom: house, key, map pin, calendar, phone, search, mail envelope, camera, folder, document, settings gear, user profile, bell, heart, star, clock, upload arrow, download arrow, share, trash bin, chat bubble, bookmark, image, plus. Construct every icon the way the logo is constructed: its main form drawn as two or three nested thin parallel lines with rounded arches and corners, at the stroke weight of the logo, and give every icon one flowing swirl stroke taken from the motif, for example as the tail of the map pin, the bell clapper, the smoke of the chat bubble or the fold of the bookmark. Coral lines on the deep purple of the kit, with the swirl stroke in periwinkle. Icons only, flat, consistent optical size across the set.", "criteria": "Interpretation of the kit first: nested-line construction and the swirl motif present across the whole set, so the icons read as the same family as the logo; then completeness and order; then consistency of the line system across twenty-four; then cleanliness. Compare with G1.", "inputs": [ "brand_kit" ], "overlay": false, "cells": [ { "key": "runchat-image", "name": "Runchat Image", "credits": 4, "score": 3, "dims": { "design": 3, "style": 4, "prompt": 3, "aesthetic": 4 }, "note": "White background, duplicated icons, several missing; the kit was not read.", "latencyMs": 4717, "method": "original blind review" }, { "key": "muse-image", "name": "Muse Image", "credits": 20, "score": 7, "dims": { "design": 8, "style": 7, "prompt": 8, "aesthetic": 7 }, "note": "Complete and ordered on purple tiles; swirls added as spiral blobs rather than integrated strokes.", "latencyMs": 22032, "method": "original blind review" }, { "key": "gpt-image-2-5-sunburst-low", "name": "GPT Image 2.5 (low)", "credits": 40, "score": 8, "dims": { "design": 8, "style": 8, "prompt": 8, "aesthetic": 8 }, "note": "All 24 requested icons appear in order in a 6-by-4 grid with the specified colours. Several main forms simplify the nested-line construction, and some periwinkle strokes read as decoration rather than an integrated part of the icon.", "latencyMs": 19798, "method": "non-blind follow-up" }, { "key": "nano-banana-1", "name": "Nano Banana 1", "credits": 80, "score": 4, "dims": { "design": 4, "style": 7, "prompt": 4, "aesthetic": 6 }, "note": "Nested lines and swirls, but phone and search swapped, user missing, a document duplicated and LOGO placeholders inside several icons.", "latencyMs": 13211, "method": "original blind review" }, { "key": "nano-banana-pro-2k", "name": "Nano Banana Pro (2K)", "credits": 300, "score": 9, "dims": { "design": 9, "style": 9, "prompt": 9, "aesthetic": 9 }, "note": "All twenty-four in order, nested double lines throughout, a periwinkle swirl integrated into every icon; the set reads as the kit.", "latencyMs": 30076, "method": "original blind review" }, { "key": "flux-2-pro", "name": "Flux 2 Pro", "credits": 100, "score": 4, "dims": { "design": 3, "style": 7, "prompt": 3, "aesthetic": 6 }, "note": "Icons framed in the logo arch, a nice reading of the kit, but many duplicates and omissions and the order lost.", "latencyMs": 18225, "method": "original blind review" } ] }, { "id": "P1-poster-basic", "capability": "poster_generate", "prompt": "Design a Swiss style poster, modern, bold and flat, using the photo in image 1. Text: TIMBER FUTURES; A lecture by Dr Mara Lindqvist; THURSDAY 21 MAY 2026, 6:30PM; MELBOURNE SCHOOL OF DESIGN, ROOM B117; FREE ENTRY; and the paragraph: Mass timber is moving from pilot projects to city blocks. This lecture traces what changed in the last five years and what it means for the next ten.", "criteria": "Text accuracy first (every string, spelled right, paragraph legible), then how convincingly it reads as Swiss style (grid, grotesk type, asymmetry, flat colour), then use of the photo and hierarchy, then overall design quality. BASIC brief: the layout is left to the model.", "inputs": [ "ref_azuma_tower" ], "overlay": false, "cells": [ { "key": "runchat-image", "name": "Runchat Image", "credits": 4, "score": 3, "dims": { "design": 3, "style": 4, "prompt": 3, "aesthetic": 5 }, "note": "Photo re-illustrated, lecturer line misspelled, venue truncated and the paragraph garbled.", "latencyMs": 5577, "method": "original blind review" }, { "key": "muse-image", "name": "Muse Image", "credits": 20, "score": 7, "dims": { "design": 8, "style": 6, "prompt": 7, "aesthetic": 8 }, "note": "All required text exact and the photo used, but an invented footer line was added and the style is more magazine than Swiss.", "latencyMs": 75938, "method": "original blind review" }, { "key": "gpt-image-2-5-sunburst-low", "name": "GPT Image 2.5 (low)", "credits": 40, "score": 7, "dims": { "design": 8, "style": 7, "prompt": 7, "aesthetic": 8 }, "note": "Required title, lecturer, event details and paragraph are legible and correctly spelled. Adds an unrequested school logo/footer and cuts out the tower instead of using a rectangular photo block; title effects also weaken the flat brief.", "latencyMs": 19197, "method": "non-blind follow-up" }, { "key": "nano-banana-1", "name": "Nano Banana 1", "credits": 80, "score": 2, "dims": { "design": 2, "style": 5, "prompt": 2, "aesthetic": 4 }, "note": "Building re-illustrated, Lindqvist and the venue misspelled, paragraph garbled.", "latencyMs": 10640, "method": "original blind review" }, { "key": "nano-banana-pro-2k", "name": "Nano Banana Pro (2K)", "credits": 300, "score": 9, "dims": { "design": 9, "style": 9, "prompt": 9, "aesthetic": 9 }, "note": "Every string and the paragraph exact, halftone photo on a grid with black and orange blocks; unmistakably Swiss. Only the prompt semicolons carried over.", "latencyMs": 27262, "method": "original blind review" }, { "key": "flux-2-pro", "name": "Flux 2 Pro", "credits": 100, "score": 5, "dims": { "design": 5, "style": 7, "prompt": 5, "aesthetic": 6 }, "note": "Photo and paragraph right, but the year was dropped, the venue line was split and duplicated, and FREE lost an E.", "latencyMs": 31350, "method": "original blind review" } ] }, { "id": "P2-poster-advanced", "capability": "poster_generate", "prompt": "Design a portrait poster in a bold modern Swiss style that pops: strict grid, a single grotesk sans-serif, flush-left text, asymmetric composition, flat colour. Take the black-and-white photo in image 1 and adjust its perspective, tilting and skewing the building into a dynamic diagonal so it drives up through the poster and gives the composition depth and movement; let it run off two edges. Use a strong palette of a saturated red, black and white, with the photo tinted into the palette. Build layering: the title \"TIMBER FUTURES\" in very large bold capitals partly behind the building and partly in front of it, so the tower cuts through the letters; a flat red block or band overlapping the photo carrying \"A lecture by Dr Mara Lindqvist\"; the paragraph \"Mass timber is moving from pilot projects to city blocks. This lecture traces what changed in the last five years and what it means for the next ten.\" in a narrow column at readable body size sitting on clear ground; and \"THURSDAY 21 MAY 2026, 6:30PM\", \"MELBOURNE SCHOOL OF DESIGN, ROOM B117\" and \"FREE ENTRY\" stacked in a small tabular block aligned to the grid. Strong hierarchy: title, then photo, then red band, then details.", "criteria": "Text accuracy first, then whether the photo was genuinely re-perspectived rather than just cropped or rotated, then the layering (type behind and in front of the building is the hard part), then palette and grid discipline, then whether it actually pops. ADVANCED brief. Compare with P1.", "inputs": [ "ref_azuma_tower" ], "overlay": false, "cells": [ { "key": "runchat-image", "name": "Runchat Image", "credits": 4, "score": 2, "dims": { "design": 2, "style": 5, "prompt": 4, "aesthetic": 4 }, "note": "Photo barely re-perspectived and most of the text garbled.", "latencyMs": 6869, "method": "original blind review" }, { "key": "muse-image", "name": "Muse Image", "credits": 20, "score": 9, "dims": { "design": 8, "style": 9, "prompt": 9, "aesthetic": 9 }, "note": "Tower tilted into a red-tinted diagonal running off two edges, FUTURES cut by the building, red and white blocks on a grid; one duplicated word in the paragraph.", "latencyMs": 106600, "method": "original blind review" }, { "key": "gpt-image-2-5-sunburst-low", "name": "GPT Image 2.5 (low)", "credits": 40, "score": 6, "dims": { "design": 7, "style": 8, "prompt": 5, "aesthetic": 6 }, "note": "Delivers a dramatic re-perspectived tower and red/black/white hierarchy, but two paragraph lines collide, making the body text unsuitable for publication. The title mostly sits in front rather than weaving behind and in front of the building.", "latencyMs": 19477, "method": "non-blind follow-up" }, { "key": "nano-banana-1", "name": "Nano Banana 1", "credits": 80, "score": 5, "dims": { "design": 5, "style": 7, "prompt": 7, "aesthetic": 6 }, "note": "Photo rotated and interleaved with the title but left black and white; paragraph and venue partly garbled.", "latencyMs": 10400, "method": "original blind review" }, { "key": "nano-banana-pro-2k", "name": "Nano Banana Pro (2K)", "credits": 300, "score": 8, "dims": { "design": 8, "style": 9, "prompt": 9, "aesthetic": 8 }, "note": "Strongest layering, the title both behind and in front of the tower, red block and grid lines; paragraph set vertically and one word garbled.", "latencyMs": 34404, "method": "original blind review" }, { "key": "flux-2-pro", "name": "Flux 2 Pro", "credits": 100, "score": 4, "dims": { "design": 3, "style": 6, "prompt": 6, "aesthetic": 5 }, "note": "Tilt and interleave present but the red block text is gibberish and the paragraph breaks down.", "latencyMs": 20816, "method": "original blind review" } ] }, { "id": "L1-relight-short", "capability": "relight_realism", "prompt": "Show the building lit with bright morning sun.", "criteria": "Did the light actually move to a convincing low morning sun with consistent shadows, first; then design and camera preservation; then how much realism the model added unprompted to trees, people and materials; then overall photographic quality. SHORT prompt.", "inputs": [ "enscape_render_red@4x3" ], "overlay": true, "cells": [ { "key": "runchat-image", "name": "Runchat Image", "credits": 4, "score": 4, "dims": { "design": 6, "style": 5, "prompt": 5, "aesthetic": 4 }, "note": "Blown-out flare from the right, fins shifted to brown, washed out.", "latencyMs": 5093, "method": "original blind review" }, { "key": "muse-image", "name": "Muse Image", "credits": 20, "score": 9, "dims": { "design": 9, "style": 9, "prompt": 9, "aesthetic": 9 }, "note": "Low sun from the left raking across the fins with flare and long shadows, photographic trees and people, design and camera kept.", "latencyMs": 99181, "method": "original blind review" }, { "key": "gpt-image-2-5-sunburst-low", "name": "GPT Image 2.5 (low)", "credits": 40, "score": 5, "dims": { "design": 8, "style": 5, "prompt": 5, "aesthetic": 6 }, "note": "Keeps the red fins and overall building but mainly increases saturation and brightness. The sun is not convincingly low morning light, and the foliage and people retain a rendered character.", "latencyMs": 20169, "method": "non-blind follow-up" }, { "key": "nano-banana-1", "name": "Nano Banana 1", "credits": 80, "score": 6, "dims": { "design": 9, "style": 5, "prompt": 5, "aesthetic": 6 }, "note": "Crisp clear daylight but the sun is high and flat; still reads as a render.", "latencyMs": 8845, "method": "original blind review" }, { "key": "nano-banana-pro-2k", "name": "Nano Banana Pro (2K)", "credits": 300, "score": 8, "dims": { "design": 9, "style": 8, "prompt": 8, "aesthetic": 9 }, "note": "Warm low sun from behind right, convincing realism pass on planting and people, design intact.", "latencyMs": 29903, "method": "original blind review" }, { "key": "flux-2-pro", "name": "Flux 2 Pro", "credits": 100, "score": 7, "dims": { "design": 8, "style": 8, "prompt": 7, "aesthetic": 8 }, "note": "Low golden sun with flare through the glazing, dramatic and plausible, sky reads a little late in the day.", "latencyMs": 15319, "method": "original blind review" } ] }, { "id": "L2-relight-detailed", "capability": "relight_realism", "prompt": "Show this building lit with bright early-morning sun: low sun from the east side raking across the red fins, crisp long shadows on the facade and ground, a clear pale-blue sky and cool shadows with warm sunlit surfaces. Produce a high-quality architectural photograph: photographic trees with real foliage and bark, real people with natural poses and clothing, real cars, believable road and paving textures, accurate reflections in the glazing, natural depth of field and a professional colour grade. Keep the building design, the red fins, the set-back upper level, the glazing and the camera and framing exactly as in the render.", "criteria": "Same as L1 with the DETAILED prompt: light direction, sky and every realism target are spelled out. Score the light first, then preservation, then how many of the realism targets were delivered, then photographic quality. Compare the pair.", "inputs": [ "enscape_render_red@4x3" ], "overlay": true, "cells": [ { "key": "runchat-image", "name": "Runchat Image", "credits": 4, "score": 7, "dims": { "design": 9, "style": 7, "prompt": 8, "aesthetic": 7 }, "note": "Low sun from the right with warm glazing; realistic but flatter shadows.", "latencyMs": 5134, "method": "original blind review" }, { "key": "muse-image", "name": "Muse Image", "credits": 20, "score": 9, "dims": { "design": 9, "style": 9, "prompt": 9, "aesthetic": 9 }, "note": "Low sun from the left with flare and long ground shadows, photographic people, bikes and cars; design and camera kept.", "latencyMs": 85320, "method": "original blind review" }, { "key": "gpt-image-2-5-sunburst-low", "name": "GPT Image 2.5 (low)", "credits": 40, "score": 7, "dims": { "design": 8, "style": 7, "prompt": 7, "aesthetic": 7 }, "note": "Warmer directional light, stronger shadows and improved surface detail, with the main design intact. Morning sun and photographic entourage are less convincing than the saved Muse and Nano Banana Pro results.", "latencyMs": 16929, "method": "non-blind follow-up" }, { "key": "nano-banana-1", "name": "Nano Banana 1", "credits": 80, "score": 6, "dims": { "design": 9, "style": 6, "prompt": 6, "aesthetic": 7 }, "note": "Crisp and clean but the sun sits high; still reads as a render.", "latencyMs": 10301, "method": "original blind review" }, { "key": "nano-banana-pro-2k", "name": "Nano Banana Pro (2K)", "credits": 300, "score": 9, "dims": { "design": 9, "style": 9, "prompt": 8, "aesthetic": 9 }, "note": "Soft raking light from behind right with tree shadows on the facade, the most photographic entourage; sun a little less low than asked.", "latencyMs": 35142, "method": "original blind review" }, { "key": "flux-2-pro", "name": "Flux 2 Pro", "credits": 100, "score": 5, "dims": { "design": 6, "style": 7, "prompt": 6, "aesthetic": 7 }, "note": "Dramatic light, but the set-back upper level was redesigned with balconies and planters and the fins turned scarlet.", "latencyMs": 19366, "method": "original blind review" } ] }, { "id": "L5-style-villa-short", "capability": "relight_realism", "prompt": "Make image 1 look like the photograph in image 2.", "criteria": "Style transfer first: does the output carry the light, grade and photographic character of image 2; then design and camera preservation; then how far the setting and planting were brought into the world of image 2; then photographic quality. SHORT prompt.", "inputs": [ "enscape_render_villa_clean@16x9", "photo_house_hillside" ], "overlay": true, "cells": [ { "key": "runchat-image", "name": "Runchat Image", "credits": 4, "score": 3, "dims": { "design": 9, "style": 2, "prompt": 3, "aesthetic": 6 }, "note": "Sunny blue sky and green lawn; the reference style was ignored.", "latencyMs": 5441, "method": "original blind review" }, { "key": "muse-image", "name": "Muse Image", "credits": 20, "score": 7, "dims": { "design": 9, "style": 7, "prompt": 7, "aesthetic": 7 }, "note": "Overcast sky and dry hills with the house and camera held; palms and grass still partly CG.", "latencyMs": 30724, "method": "original blind review" }, { "key": "gpt-image-2-5-sunburst-low", "name": "GPT Image 2.5 (low)", "credits": 40, "score": 7, "dims": { "design": 8, "style": 7, "prompt": 7, "aesthetic": 8 }, "note": "Transfers muted overcast light and improves the vegetation and materials while largely keeping the house and viewpoint. Retains prominent palms and does not fully adopt the reference's dry native landscape.", "latencyMs": 19453, "method": "non-blind follow-up" }, { "key": "nano-banana-1", "name": "Nano Banana 1", "credits": 80, "score": 6, "dims": { "design": 5, "style": 8, "prompt": 7, "aesthetic": 7 }, "note": "Reference mood carried well, but the viewpoint moved and the house warped.", "latencyMs": 10781, "method": "original blind review" }, { "key": "nano-banana-pro-2k", "name": "Nano Banana Pro (2K)", "credits": 300, "score": 6, "dims": { "design": 5, "style": 9, "prompt": 7, "aesthetic": 9 }, "note": "Fog, dry grass hills and dead trees under a flat sky with a muted grade, but the viewpoint moved: the house sits differently in the frame. (Moved down on review.)", "latencyMs": 34357, "method": "original blind review" }, { "key": "flux-2-pro", "name": "Flux 2 Pro", "credits": 100, "score": 8, "dims": { "design": 6, "style": 9, "prompt": 8, "aesthetic": 9 }, "note": "Closest tone and grade to the reference photograph; the camera dropped and the ground was rebuilt as a bank. (Moved up on review.)", "latencyMs": 15991, "method": "original blind review" } ] }, { "id": "L6-style-villa-detailed", "capability": "relight_realism", "prompt": "Edit the render in image 1 so that it matches the realism and photographic style of image 2: the same soft overcast light with no hard shadows, the same muted, slightly desaturated colour grade with lifted blacks, the same cool neutral sky, the same lens feel and depth. Replace the rendered planting and lawn with photographic planting in the character of image 2, dry native grasses and eucalypt-like trees, and make the paving, timber, glazing reflections and car photographic. Keep the house design, the cladding, the glazing, the pergola, the terrace and the camera and framing exactly as in image 1.", "criteria": "Same as L5 with the DETAILED prompt: every element of the style and every realism target is spelled out. Score the style match first, then preservation, then delivered targets, then photographic quality. Compare the pair.", "inputs": [ "enscape_render_villa_clean@16x9", "photo_house_hillside" ], "overlay": true, "cells": [ { "key": "runchat-image", "name": "Runchat Image", "credits": 4, "score": 7, "dims": { "design": 9, "style": 7, "prompt": 7, "aesthetic": 7 }, "note": "Muted overcast grade with dry grasses, house held, but the palms stayed and the image is flat.", "latencyMs": 6159, "method": "original blind review" }, { "key": "muse-image", "name": "Muse Image", "credits": 20, "score": 9, "dims": { "design": 9, "style": 9, "prompt": 9, "aesthetic": 9 }, "note": "Overcast light, lifted blacks, eucalypts and dry grasses with hills behind; house and camera kept; reads as the reference series.", "latencyMs": 27839, "method": "original blind review" }, { "key": "gpt-image-2-5-sunburst-low", "name": "GPT Image 2.5 (low)", "credits": 40, "score": 8, "dims": { "design": 7, "style": 9, "prompt": 8, "aesthetic": 9 }, "note": "Convincing muted overcast photograph with dry grasses and eucalypt-like trees; house and pergola remain recognisable. Replaces much of the patterned grass-paved drive with dry ground, so preservation is incomplete.", "latencyMs": 19390, "method": "non-blind follow-up" }, { "key": "nano-banana-1", "name": "Nano Banana 1", "credits": 80, "score": 5, "dims": { "design": 9, "style": 6, "prompt": 6, "aesthetic": 5 }, "note": "Overcast and dry grass applied over a still-CG house; soft.", "latencyMs": 11132, "method": "original blind review" }, { "key": "nano-banana-pro-2k", "name": "Nano Banana Pro (2K)", "credits": 300, "score": 6, "dims": { "design": 6, "style": 9, "prompt": 7, "aesthetic": 8 }, "note": "Very photographic and on-mood, but the cladding and upper level were redesigned.", "latencyMs": 31939, "method": "original blind review" }, { "key": "flux-2-pro", "name": "Flux 2 Pro", "credits": 100, "score": 9, "dims": { "design": 8, "style": 9, "prompt": 8, "aesthetic": 8 }, "note": "Reference mood and planting delivered; the foreground paving became a grass bank. (Moved up on review: tone closest to the reference.)", "latencyMs": 16137, "method": "original blind review" } ] } ] }