Head to head: Seedream 5.5 Layerize vs OpenAI: GPT-5.4 Image 2
Seedream 5.5 Layerize vs OpenAI: GPT-5.4 Image 2
By Ryan Merket · Published
A close image-model matchup across cinematic scenes, constrained palettes, exact counts, and text-heavy designs. The comparison turns on how each handles specific prompt requirements, not on a single kind of image.
Seedream 5.5 Layerize won four tasks: rainy showroom teardown, perspective & scale, palette-locked field lab, and exact counting. Its strongest pattern was meeting composition and framing requirements, particularly in the 16:9 scene briefs and the seven-cup count. OpenAI: GPT-5.4 Image 2 took three tasks—macro cable braid bench, legible multi-line text, and Memphis repair kiosk—while restricted palette ended in a tie. The split leaves meaningful wins on both sides rather than a decisive advantage across the varied briefs. The aggregate scores were **60.3** for Seedream and **63.0** for OpenAI, but the saved result is a tie. **Too close to call (not statistical): this assessment applies only to the saved task sample, and it does not establish a general capability advantage for either model.**
Rainy showroom teardown
A photorealistic evening scene inside a boutique e-bike showroom during a rainstorm, viewed in cinematic 16:9 from just outside the front glass as a technician at an oak worktable opens the battery housing of a matte graphite commuter bike; wet pavement reflects the window light, raindrops bead and streak on the glass, brushed aluminum tools catch cool highlights, black rubber tires absorb light differently from lacquered wood shelving, skin tones, woven jacket fabric, tempered glass, anodized metal, and paper manuals all need physically believable material response under mixed warm pendant lighting and blue streetlight spill.
OpenAI: GPT-6 Luna: Model A more fully matches the cinematic 16:9 view through rain-streaked glass, with the technician opening the battery housing at a well-equipped oak worktable and convincing warm/cool material lighting. Model B captures the rainy showroom and repair, but its square framing, less prominent worktable, and distracting signage make it a weaker fit. (Second judge pass, order swapped — scores are the average of both: Model A more closely matches the cinematic 16:9 view and clearly shows the technici | Anthropic: Claude Fable 5.1: Model A delivers the requested 16:9 cinematic f…
Restricted palette
A flat-vector illustration of a mountain campsite at dusk using ONLY four colors — cream, burnt orange, teal, and charcoal — with no gradients and no stray off-palette colors, balanced composition, 16:9.
OpenAI: GPT-6 Luna: Model A offers a polished, balanced 16:9 campsite scene with the requested mountain-and-dusk setting, though its many tonal variations soften the strict four-color constraint. Model B uses a cleaner restricted-palette vector style, but its square format notably misses the requested aspect ratio and its composition feels less complete. (Second judge pass, order swapped — scores are the average of both: Model B more closely follows the flat-vector style and limited four-color palette, though its s | Anthropic: Claude Fable 5.1: Model A delivers the 16:9 frame and a richly de…
Perspective & scale
A photorealistic one-point-perspective shot down a long empty library aisle, shelves receding to a single vanishing point, books getting consistently smaller with distance and nothing warped or wrongly sized, warm overhead light, 16:9.
OpenAI: GPT-6 Luna: Model A more fully matches the requested 16:9 framing while presenting a clear, realistic one-point-perspective aisle with warm overhead lighting. Model B has strong depth and scale, but its nearly square framing misses the specified aspect ratio. (Second judge pass, order swapped — scores are the average of both: Model A more closely matches the requested 16:9 photorealistic aisle, with convincing scale progression and a clear central vanishing point. Model B has a particularly clean, symmetric | Anthropic: Claude Fable 5.1: Model A delivers a true 16:9 frame with clean o…
Macro cable braid bench
An extreme close-up of a watchmaker-like electronics bench where a custom mechanical keyboard cable is being hand-braided beside a disassembled titanium smart ring, with crisp macro focus on fine nylon weave, individual copper strands, brushed tool steel, knurled screwdriver handles, dust motes, fingerprint smears on a magnifier lens, laser-etched serial text, and micro-scratches on anodized parts; the composition should feel intimate and precise, with every tiny texture sharply resolved and no waxy smudging or melted detail.
OpenAI: GPT-6 Luna: Model A clearly captures the cable being hand-braided at a working electronics bench, while Model B more fully resolves the requested weave, exposed copper strands, disassembled ring, etched text, tools, and lens grime. Model B’s still-life arrangement is less intimate and lacks a visible hand, but its sharper macro detail and overall prompt coverage make it the stronger result. (Second judge pass, order swapped — scores are the average of both: Model B more fully delivers the extreme macro brie | Anthropic: Claude Fable 5.1: Model B delivers the macro intimacy the prompt…
Palette-locked field lab
A compact autonomous soil-analysis rover parked beside a fold-out workstation in a windswept salt marsh at dusk, shown in a clean 16:9 side-angle composition; restrict the entire image strictly to the named Desert Signal palette only: adobe red, clay pink, sun-bleached sand, sage green, oxidized teal, midnight indigo, and warm ivory, with no other hues anywhere in the scene, including sky, reflections, screens, cables, and shadows; the rover, sample vials, antenna mast, marsh grasses, distant pylons, and cloud bands must all stay inside that palette while remaining clearly distinguishable and visually rich.
OpenAI: GPT-6 Luna: Model A better matches the requested 16:9 side-angle composition and makes the Desert Signal palette vivid and cohesive, though it adds a second rover. Model B depicts the workstation and marsh convincingly, but its square format misses the requested aspect ratio and its palette is less strictly controlled. (Second judge pass, order swapped — scores are the average of both: Model A better matches the requested 16:9 side-angle composition and uses the restricted palette with strong visual clarity | Anthropic: Claude Fable 5.1: Model A delivers the requested 16:9 side-angle…
Exact counting
A clean overhead flat lay on white marble of EXACTLY seven identical espresso cups arranged in a single straight row, each clearly separate and individually countable, even soft lighting, 16:9.
OpenAI: GPT-6 Luna: Model A shows exactly seven clearly separated cups in a straight row and matches the requested 16:9 overhead marble presentation. Model B also has seven countable cups, but its square format misses the requested aspect ratio. (Second judge pass, order swapped — scores are the average of both: Model B shows exactly seven clearly separated cups in a clean overhead row, but its square framing misses the requested 16:9 format and leaves the cups relatively small. Model A also has exactly seven disti | Anthropic: Claude Fable 5.1: Model A delivers exactly seven identical brown…
Legible multi-line text
A minimalist event poster with three lines of crisp, correctly-spelled text stacked and centered: 'NIGHT MARKET' large on top, 'Fridays · 6–11pm' in the middle, 'Riverside Pier 4' at the bottom, on a deep navy background, clean sans-serif, subtle grain.
OpenAI: GPT-6 Luna: Model A presents the exact three requested lines, centered and legible, with a cohesive nighttime riverside scene that supports the event theme. Model B has polished typography, but breaks “NIGHT MARKET” into two lines, so it does not follow the requested three-line layout. (Second judge pass, order swapped — scores are the average of both: Model A presents the requested wording in exactly three centered lines with crisp, correctly spelled text, while Model B breaks the title across two lines. M | Anthropic: Claude Fable 5.1: Model B delivers a truly minimalist poster with…
Memphis repair kiosk
A sidewalk micro-kiosk called Circuit Canteen where a street technician is repairing translucent wireless earbuds under a striped awning, depicted faithfully in 1980s Memphis design style with bold squiggles, terrazzo-like pattern blocks, playful geometric forms, asymmetry, and candy-colored flat shapes; include stacked battery packs, oversized receipt tape curling onto the pavement, a tiny QR sign, and a diagonally framed 16:9 composition lit by bright noon sun, all rendered so the image unmistakably reads as Memphis rather than generic retro illustration.
OpenAI: GPT-6 Luna: Model A delivers the stronger unmistakable Memphis look, with vivid asymmetrical geometric forms, terrazzo-like blocks, a clear kiosk title, and the requested 16:9 sidewalk scene. Model B includes the repair activity and accessories more explicitly, but its squarer, more realistic presentation feels less compositionally faithful to the prompt. (Second judge pass, order swapped — scores are the average of both: Model A more unmistakably delivers the requested Memphis look, with vivid flat shapes, | Anthropic: Claude Fable 5.1: Model A delivers the requested 16:9 flat candy-…
Matchup powered by OpenRouter.