Head to head: Ideogram V4.5 Edit vs Bytedance Seedream V4.5 Edit
Ideogram V4.5 Edit vs Bytedance Seedream V4.5 Edit
By Ryan Merket · Published
This edit-model matchup tests more than photorealism: it asks whether each system can hold a composition together while obeying narrow style, palette, and content constraints. Across the prompts, the trade-off is often between a polished overall scene and a particularly strong detail or material rendering.
Ideogram V4.5 Edit takes the saved verdict, with aggregate scores of **61.3** to Bytedance Seedream V4.5 Edit’s **56.3**. It won five tasks to Seedream’s three, with no ties—a **slight edge (not statistical)** on this saved task sample, not a runaway result. Ideogram’s advantage is most visible when the prompt demands disciplined art direction. It came out ahead on the ukiyo-e scene, Bauhaus poster, restricted-palette campsite, and melancholy kiosk, where composition and fidelity to the requested visual language mattered as much as including the right objects. Its reading nook also felt more complete and balanced while respecting the exclusions. Seedream’s three wins came in the bound-gadget still life, perspective-and-scale aisle, and repair-bench scene. The still life showed the value of its material rendering, while the empty aisle better honored a key scene constraint. These results keep the matchup close: Seedream found meaningful wins, even as Ideogram more consistently delivered the requested composition and style across the sample. **Final call: Ideogram V4.5 Edit wins this head-to-head by a slight editorial edge. The assessment applies only to the saved task sample.**
Negation
A cozy reading nook with an armchair, a stack of books, and a mug of tea by a window — with absolutely NO plants, NO lamps, and NO artwork or picture frames anywhere in the frame. Warm afternoon light, 16:9.
OpenAI: GPT-6 Luna: Model A presents a complete, inviting reading nook with the requested chair, book stack, tea, and afternoon window light, while keeping the prohibited items out of frame. Model B also avoids plants, lamps, and artwork, but its very tight framing and intensely saturated light make the nook feel less balanced and complete. (Second judge pass, order swapped — scores are the average of both: Model B captures the requested warm, cozy afternoon mood especially well and includes the chair, books, tea, | Anthropic: Claude Fable 5.1: Model A delivers a polished, cozy nook with armc…
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 more faithfully captures ukiyo-e woodblock conventions, with a traditional fishing boat, bold contours, stylized foam, and a muted indigo-and-cream palette. Model B has an attractive dawn scene, but its large modern-looking vessel and more diffuse, expansive composition are less aligned with the prompt. (Second judge pass, order swapped — scores are the average of both: Model A more convincingly captures ukiyo-e through its bold flat forms, stylized curling foam, and indigo-and-cream pal | Anthropic: Claude Fable 5.1: Model A delivers a convincingly traditional uki…
Bound gadget still life
Ultra-clean studio product still life, 16:9: five separate tech objects arranged on a pale concrete slab, each with its own exact color and material that must remain distinct — a matte teal ceramic smart mug, a brushed brass pocket SSD, a translucent violet acrylic game controller, a glossy coral glass webcam, and a slate-black carbon-fiber stylus; include a small folded card with tiny gray calibration marks but no other colored props; soft side lighting from the left, faint shadows, high detail, colors locked to the correct objects with no bleeding, swapping, or mixed materials, minimalist composition.
OpenAI: GPT-6 Luna: Model A presents all five objects and the marked card clearly in a balanced, minimalist still life, while Model B crops several objects heavily and lets the mug dominate the frame. Model B renders the brass SSD and carbon-fiber stylus convincingly, but Model A better fulfills the overall composition and presentation requested. (Second judge pass, order swapped — scores are the average of both: Model A presents all five distinct objects and the calibration card clearly in a balanced, minimalist s | Anthropic: Claude Fable 5.1: Model B renders each material unmistakably — vi…
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 has a convincing centered vanishing point, coherent shelf scale, and a more polished photorealistic atmosphere, though the book on the floor conflicts with the empty-aisle request. Model B keeps the aisle empty and uses warm overhead lighting, but its strong softness and somewhat exaggerated foreground perspective make it feel less realistic. (Second judge pass, order swapped — scores are the average of both: Model B more cleanly delivers the requested empty, warmly lit aisle with strong | Anthropic: Claude Fable 5.1: Model B delivers a cleanly converging, truly em…
Melancholy kiosk vector
Clean vector illustration, 16:9: a lonely midnight electronics kiosk on a rain-slick tram platform selling obscure gadgets, with one illuminated shelf displaying a moon-phase alarm clock, a foldable e-paper map reader, and a boxy pager-like translator; no people present, only the suggestion of recent activity from a tipped receipt and a half-closed shutter; mood is quiet, wistful, slightly futuristic; use simplified geometric shapes, crisp outlines, restrained palette of deep indigo, foggy mint, and sodium-amber highlights; strong composition with the kiosk centered and reflections stretching across wet tiles.
OpenAI: GPT-6 Luna: Model A more closely matches the clean geometric vector style and restrained, wistful composition, with a centered kiosk, half-closed shutter, tipped receipt, and clear reflections across the platform. Model B includes the requested merchandise and atmospheric wet station, but its painterly detail and close crop depart more from the specified vector look. (Second judge pass, order swapped — scores are the average of both: Model B captures the rain, reflections, and illuminated gadgets beautifull | Anthropic: Claude Fable 5.1: Model A delivers a true clean vector look with…
Bauhaus audio poster
Faithful Bauhaus style poster image, 16:9: a front-facing modular smart speaker tower presented as a design exhibition advertisement, built from stacked circles, rectangles, and semicircles with visible grille patterns reduced to bold geometry; use a disciplined Bauhaus palette of cream, black, brick red, mustard yellow, and muted blue; asymmetrical but balanced layout, sans-serif typographic blocks as abstract placeholders, flat colors, sharp edges, generous negative space, and the unmistakable functional modernist character of 1920s Bauhaus graphic design.
OpenAI: GPT-6 Luna: Model A closely matches the requested flat, disciplined Bauhaus poster, with a clear modular speaker, balanced negative space, fitting palette, and legible typography. Model B includes the speaker and palette, but its room-like rendering, cropped elements, cluttered layout, and distorted oversized text depart substantially from the prompt. (Second judge pass, order swapped — scores are the average of both: Model A has a disciplined Bauhaus palette, crisp geometric forms, generous negative space, | Anthropic: Claude Fable 5.1: Model A delivers a genuinely flat, asymmetrical…
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 presents a clear, balanced campsite with mountains, tent, fire, and a cohesive four-color flat-vector palette. Model B uses the requested colors, but the oversized, tightly cropped tent and mountain overwhelm the scene, and its visible texture weakens the flat, no-gradient style. (Second judge pass, order swapped — scores are the average of both: Model A clearly depicts a balanced mountain campsite at dusk in a clean flat-vector style, while Model B is dominated by an oversized mountain | Anthropic: Claude Fable 5.1: Model A stays much closer to the strict four-col…
Repair bench hands
Cinematic editorial photo, 16:9: a robotics repair technician with natural adult proportions leans over a cluttered workbench in a compact satellite-lab, clearly visible from head to mid-thigh, calmly replacing a tiny cooling fan in a transparent cobalt drone controller; both hands are fully in frame and anatomically correct, with all ten fingers visible and plausible, the left hand pinching a precision screwdriver while the right hand steadies a ribbon cable, no extra fingers, no fused fingers, no warped wrists; short copper hair, rolled charcoal coveralls, subtle grease smudge on one forearm, shelves of labeled parts drawers behind; cool overhead strip lights mixed with a warm desk lamp, crisp depth of field, realistic skin texture and believable pose.
OpenAI: GPT-6 Luna: Model A closely matches the requested repair scene, with the technician, transparent cobalt device, screwdriver, cable, labeled drawers, and both hands clearly engaged in the work. Model B has appealing cinematic lighting and a visible drone fan, but the close crop cuts off the hands and misses the requested head-to-mid-thigh framing. (Second judge pass, order swapped — scores are the average of both: Model A better matches the requested full, readable repair scene, with both hands visible worki | Anthropic: Claude Fable 5.1: Model A delivers a convincingly photographic sc…
Matchup powered by OpenRouter.