Head to head: Ideogram V4.5 Edit vs Krea 3 Edit
Ideogram V4.5 Edit vs Krea 3 Edit
By Ryan Merket · Published
This matchup tests more than visual polish: the prompts demand legible lettering, exact counts, convincing hands, and fidelity to specific scenes and styles. The verdict turns on which model handles that varied mix more consistently in the saved task sample.
Krea 3 Edit takes the saved result, 5 task wins to Ideogram V4.5 Edit’s 3, with no ties. The aggregate scores are **61.9 to 58.9**—a close margin, not a blowout. Krea’s clearest wins came on the white “POLAR SPROUTS” crate and the row of exactly seven cups, where clean presentation and legibility mattered as much as the subject itself. Ideogram had the better showing on hands and anatomy, the inventory scene, and the named art-style prompt. Its craft scene made the hands’ work on the bracelet easier to read; its inventory image was closer to the requested supply-room scene, though the count was still uncertain; and its fishing boat was more visible against the dawn wave. The remaining prompts underline why this is not a sweeping verdict: Krea won the Bauhaus toolmakers, neon dispatch, and restricted-palette tasks in the saved results, but the images were not uniformly prompt-perfect. In particular, both palette attempts showed color variation beyond a strict flat-color reading. This assessment applies to the saved task sample, not to every use of either model. **Final call: Krea 3 Edit, with a slight edge—not a statistical one.**
Hands & anatomy
A natural candid photo of a person tying a friendship bracelet, both hands clearly visible with anatomically correct fingers and a plausible grip on the threads, soft daylight, shallow depth of field, 16:9.
OpenAI: GPT-6 Luna: Model A more clearly shows both hands naturally working the bracelet threads, with convincing finger anatomy and a candid craft setting. Model B reads as tying a bracelet onto a wrist, but the close, slightly awkward hand arrangement is less clear and natural. (Second judge pass, order swapped — scores are the average of both: Model A shows both hands clearly holding and working the bracelet, with a more plausible grip and a natural, softly lit crafting scene. Model B includes the bracelet on a | Anthropic: Claude Fable 5.1: Model A delivers a convincing candid scene with…
Readable Crate Label
A clean editorial product-style image of a warehouse supervisor holding a white shipping crate toward the camera, with a bold printed label that is perfectly legible and reads exactly: "POLAR SPROUTS"; sans-serif typography, high contrast black text on matte white surface, minimalist line-art-inspired composition with ample negative space, soft studio lighting, sharp focus on the label.
OpenAI: GPT-6 Luna: Model B better matches the white-crate, clean product-style brief, with large, perfectly legible “POLAR SPROUTS” text and a polished studio presentation. Model A includes a clearer warehouse setting, but its crate is wood-toned rather than white and the overall image is less aligned with the requested studio aesthetic. (Second judge pass, order swapped — scores are the average of both: Model B closely follows the clean editorial brief, with a white crate and the exact, sharply legible “POLAR SPR | Anthropic: Claude Fable 5.1: Model B delivers a true white matte crate held…
Inventory Count
A documentary-style 16:9 scene inside a marine-biology supply room where a logistics technician in teal coveralls kneels beside metal shelving, with EXACTLY 13 distinct orange sample coolers arranged across the floor and lower shelf, every cooler fully visible, separated, and individually countable; overhead industrial light, crisp focus, realistic textures, eye-level composition, no extra coolers or duplicate objects obscuring the count.
OpenAI: GPT-6 Luna: Model A is closer to the requested inventory scene, though several coolers are partly obscured and the count does not clearly reach exactly 13. Model B makes the coolers easy to distinguish but appears to include substantially more than 13. (Second judge pass, order swapped — scores are the average of both: Model B has a clear, well-separated arrangement, but it appears to include substantially more than 13 coolers. Model A better conveys a marine-biology supply room and has a polished documenta | Anthropic: Claude Fable 5.1: Model A shows roughly 11–12 orange coolers (clo…
Bauhaus Toolmakers
A 16:9 workshop scene of two instrument makers calibrating a strange brass surveying device, rendered faithfully in Bauhaus style with strict geometric forms, primary-color accents, asymmetrical balance, flat planes, and disciplined modernist poster aesthetics; bright diffuse daylight from tall factory windows, simplified yet purposeful composition, every shape and color choice reinforcing authentic Bauhaus design principles.
OpenAI: GPT-6 Luna: Model A clearly shows two human instrument makers calibrating brass surveying equipment in a disciplined Bauhaus poster style, with tall factory windows and restrained primary-color accents. Model B has a compelling geometric composition, but its robot-like figures do not fit the prompt’s workshop scene as well. (Second judge pass, order swapped — scores are the average of both: Model A clearly depicts two human instrument makers working with brass surveying instruments, using a disciplined Bauh | Anthropic: Claude Fable 5.1: Model B channels an authentically Bauhaus (Schl…
Neon Night Dispatch
A 16:9 cinematic view of a bicycle courier dispatcher working alone at a cramped corner desk in a rain-slick alley office at neon night, magenta and cyan signage outside casting saturated reflections across the window and keyboard, the mood tense but energetic; realistic urban style, strong contrast, moody shadows, three-quarter view, detailed wet glass and glowing street haze.
OpenAI: GPT-6 Luna: Model A more fully conveys a lone bicycle courier dispatcher in a cramped alley office, with the courier bag, route maps, rainy neon street, and cinematic three-quarter composition all clearly integrated. Model B has vivid magenta-cyan reflections and convincing dispatch equipment, but its higher-angle view and less distinct alley-office context make it slightly less aligned with the prompt. (Second judge pass, order swapped — scores are the average of both: Model B captures the magenta-cyan ref | Anthropic: Claude Fable 5.1: Model A delivers the requested three-quarter ci…
Named art style
A ukiyo-e woodblock print of a fishing boat riding a large cresting wave at dawn, faithful to the flat color planes, bold outlines, and stylized foam of the tradition, muted indigo and cream palette.
OpenAI: GPT-6 Luna: Model A clearly depicts a fishing boat riding a dramatic cresting wave at dawn, with muted indigo and cream colors, bold outlines, and stylized foam. Model B has a compelling ukiyo-e composition, but the boat is mostly obscured and the dawn setting is not apparent. (Second judge pass, order swapped — scores are the average of both: Model A more fully depicts a fishing boat riding a dramatic cresting wave at dawn, with a visible rising sun, stylized foam, and convincing ukiyo-e linework. Model B | Anthropic: Claude Fable 5.1: Model A delivers every prompt element—a fishing…
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 B presents a more evocative dusk campsite, with a crescent moon, stars, layered mountains, trees, tent, and fire in a cohesive composition. Both images appear to use subtle gradients or extra color variations despite the strict palette request, though Model A’s oversized mountains and pale background make its scene feel less atmospheric. (Second judge pass, order swapped — scores are the average of both: Model B more clearly conveys dusk with a crescent moon and uses a tighter, more consis | Anthropic: Claude Fable 5.1: Model A holds strictly to the four flat colors…
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 cups in a straight row, but the saucers and spoons add visual clutter beyond the requested simple arrangement. Model B presents exactly seven clearly separated, identical-looking cups in a clean, evenly spaced row on white marble, so it adheres more closely to the prompt. (Second judge pass, order swapped — scores are the average of both: Model B shows exactly seven clearly separate, matching espresso cups in a clean straight row and adheres closely to the requested f | Anthropic: Claude Fable 5.1: Both Model A and Model B deliver exactly seven…
Matchup powered by OpenRouter.