Head to head: Bytedance Seedream V4.5 Edit vs Ideogram V4.5 Edit
Bytedance Seedream V4.5 Edit vs Ideogram V4.5 Edit
By Ryan Merket · Published
A set of eight image-editing tests pits precise instruction-following against polished, readable composition. The results turn on which model more consistently makes the requested action and visual constraints unmistakable.
Seedream takes the matchup **60.3 to 55.8**, with the statistical verdict putting its win at **limited confidence**. It won four tasks to Ideogram’s one, with three ties—not a sweep, but a consistent edge across this test set. The clearest advantage came when the image needed to communicate a specific action or effect. Seedream’s bracelet image showed loose threads being actively tied, and its sneaker ad made the soda splash feel more kinetic, with arcing droplets and motion in the laces. Ideogram often offered the more composed frame, but in those cases the action was less evident or the splash more restrained. The result was not uniform. Ideogram won the palette-locked tea-tin task with four distinct tins in a clean studio arrangement, while Seedream’s five tins and aggressive crop missed the brief. The lamp-assembly, freezer-label, and room-layout comparisons were ties: each model had a real strength, whether clearer assembly action, more legible text, or more faithful spatial arrangement, without a decisive overall advantage. **Final call: Bytedance Seedream V4.5 Edit wins on the strength of more task wins and a higher aggregate score. Ideogram V4.5 Edit remains competitive—and takes the tea-tin brief outright—but its cleaner presentation does not consistently compensate when the requested action or effect needs to read at a glance.**
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 shows the threads being worked, but the extreme close crop and overlapping fingers make the grip and hand anatomy less clear. Model B has more natural-looking hands, a clearer composition, and soft daylight, though it appears to show a nearly finished bracelet rather than active tying. (Second judge pass, order swapped — scores are the average of both: Model A more clearly shows the hands working loose threads to tie the bracelet, though its extreme close crop and warm lighting depart fr | Anthropic: Claude Fable 5.1: Model A shows a convincing mid-knot tying actio…
Attribute binding
A photorealistic tabletop scene: a matte green cube to the LEFT of a glossy red sphere, a small blue cylinder BEHIND both, and a single yellow rubber duck sitting ON TOP of the cube. Colors must stay on the correct objects. Soft studio light, 16:9.
OpenAI: GPT-6 Luna: Model B presents all four objects clearly in the requested left-right and front-back arrangement, with the duck visibly sitting on the cube. Model A follows the object/color instructions too, but its oversized cube and duck, partly obscured objects, and excessive empty space make the composition less effective. (Second judge pass, order swapped — scores are the average of both: Model B clearly shows all four correctly colored objects in the requested left/right and front/back arrangement, with t | Anthropic: Claude Fable 5.1: Both models bind all colors correctly and place…
Hands assembling lamp
A realistic lifestyle image of a young adult sitting at a wooden table assembling a compact mushroom-shaped desk lamp from a consumer electronics kit, shown from the waist up with both hands clearly visible in the foreground tightening a small knurled ring onto the lamp stem; anatomically correct hands with five natural fingers each, accurate thumb placement, believable joints, and natural body proportions; Scandinavian home interior, soft morning window light, candid documentary photography style, shallow depth of field, 16:9.
OpenAI: GPT-6 Luna: Model A more clearly shows both hands tightening the knurled ring on the lamp stem, with the assembly action prominent in the foreground. Model B has a more polished lifestyle composition and compact lamp, but the ring-tightening action is less clear and the hands are comparatively small in frame. (Second judge pass, order swapped — scores are the average of both: Model B better captures the requested compact lamp, kit, wooden-table setting, and candid lifestyle framing. Model A shows the knurle | Anthropic: Claude Fable 5.1: Model A delivers the requested foreground close…
Palette-locked tea tins
A pristine studio product photo of four cylindrical tea tins arranged in a shallow arc on a matte countertop, each tin decorated with abstract leaf motifs and small geometric badges, rendered ONLY in the named color palette 'Arctic Tram' with no other colors present: glacial white, coal black, petrol teal, mustard ochre, and dusty rose; no gradients outside those hues, no metallic reflections introducing extra colors, no stray off-palette pixels anywhere; crisp softbox lighting, high detail, minimalist commercial packaging style, 16:9.
OpenAI: GPT-6 Luna: Model B shows exactly four tins in a clean commercial setup and keeps the palette largely on brief, though the tins are mostly white and arranged in a row rather than an arc. Model A uses the palette more vividly, but shows five tins and crops them dramatically, weakening adherence and composition. (Second judge pass, order swapped — scores are the average of both: Model B presents four clearly separated tins in a restrained studio arrangement with recognizable palette accents, while Model A is | Anthropic: Claude Fable 5.1: Model A is far more detailed and attractive, wit…
Mid-splash sneaker ad
A dynamic advertising image of a translucent lime-green running shoe bursting through a shallow puddle of sparkling citrus soda, captured mid-stride with droplets and fizzy spray arcing backward, the sole visibly flexed and the laces whipping in motion, dramatic low-angle composition, high-speed studio sports photography look, hard rim lighting plus a cool backlight to emphasize motion trails and liquid texture, energetic and convincing sense of forward movement, 16:9.
OpenAI: GPT-6 Luna: Model A delivers the more energetic soda-splash moment, with prominent arcing droplets, motion-whipped laces, and a translucent lime-green shoe, though the extreme close crop makes the shoe somewhat distorted. Model B presents a cleaner, more legible product shot, but its splash and sense of motion are comparatively subdued. (Second judge pass, order swapped — scores are the average of both: Model B presents a polished, clearly readable sneaker, but the splash is relatively restrained and the sh | Anthropic: Claude Fable 5.1: Model A delivers a genuinely translucent lime s…
Readable freezer label
A close-up commercial product shot of a frosted glass meal-prep container with a clean adhesive label centered perfectly on the lid, the label text must be sharp, correctly spelled, and easy to read: 'NORI SOUP' on the first line and 'LOT 42B' on the second line; modern sans-serif typography, dark ink on a pale cream label, subtle condensation around the edges but not obscuring text, realistic supermarket packaging photography, bright diffused overhead lighting, 16:9.
OpenAI: GPT-6 Luna: Model A has an attractive frosted-container close-up, but the oversized, skewed lettering—especially the second line—looks distorted and partly clipped. Model B is less dramatic and the label is small, but both requested lines are correctly spelled and clearly readable on a centered lid label. (Second judge pass, order swapped — scores are the average of both: Model B shows the full container with the correctly spelled, centered two-line label, though the text is relatively small. Model A has bo | Anthropic: Claude Fable 5.1: Model A delivers a true close-up with large, sh…
Spatial layout
A clean isometric illustration of a bedroom: a bed against the LEFT wall, a round rug centered on the floor, a desk under the WINDOW on the back wall, and a floor lamp in the FRONT-RIGHT corner. Flat-vector style, consistent perspective.
OpenAI: GPT-6 Luna: Model B presents the requested furniture arrangement in a clearer, more consistent isometric layout, while Model A’s room reads more like a close, frontal perspective and is heavily cropped. Both are attractive flat-vector illustrations, though Model B’s lamp is not quite in the front-right corner. (Second judge pass, order swapped — scores are the average of both: Model B clearly follows the requested isometric layout, with the bed along the left wall, a centered round rug, a desk beneath the b | Anthropic: Claude Fable 5.1: Model A places all four items correctly (bed on…
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 closely follows the empty-aisle requirement, with shelves and books receding toward a clear central vanishing point and warm overhead lighting. Model B has a strong, atmospheric perspective, but the ladder and open book make the aisle less empty. (Second judge pass, order swapped — scores are the average of both: Model B more convincingly delivers a long, centered one-point-perspective aisle, with consistently receding shelves and warm overhead lighting. Model A has a clear central | Anthropic: Claude Fable 5.1: Model A delivers a clean, empty aisle with stric…
Matchup powered by OpenRouter.