GPT Image 1.5 vs Seedream 4.0
Head-to-head across 16 challenges
GPT Image 1.5
74.1%
win rate
Ties
0.0%
Seedream 4.0
25.9%
win rate
Challenge Results
Candid Street Photography
Text-to-Image“A candid street photo of an elderly Japanese man repairing a red bicycle in light rain, reflections on wet pavement, shallow depth of field, 50mm lens, natural skin texture, imperfect framing, motion blur from passing cars, cinematic but realistic, no stylization.”
AI Judge Analysis
GPT Image 1.5
- + Excellent attention to detail with realistic rain droplets on the man's jacket and the bike frame.
- + Strong atmospheric lighting and reflections that accurately convey a rainy day.
- + Successful execution of the 'imperfect framing' prompt with the cropped car and candid feel.
- − The bike anatomy becomes a bit messy around the rear wheel and spokes.
Seedream 4.0
- + Good use of motion blur on the passing vehicle to create a sense of street movement.
- + Realistic puddles and reflections on the pavement.
- + Clearer view of the man's face and hands working on the bike.
- − The man is wearing a short-sleeved shirt in what appears to be rainy weather, which feels logically inconsistent.
- − The bicycle's physical structure is significantly warped, particularly the pedals/crankset and the front wheel spokes.
Verdict: GPT Image 1.5 is the superior image as it better captures the textures and atmosphere of a rainy day, including visible raindrops on surfaces. Seedream 4.0 captures the motion blur well, but the physical distortions of the bicycle and the illogical clothing choice for the weather make it less convincing.
Modern Clean Menu
Text-to-Image“Modern minimalist restaurant menu design, white background with colorful food photos in grid, sections for appetizers/pizza/mains, bold sans-serif fonts, vibrant accents, clean professional layout for casual dining.”
AI Judge Analysis
GPT Image 1.5
- + Excellent text rendering with clear, legible names, descriptions, and prices.
- + Highly professional layout that functions as an actual usable menu.
- + Perfect adherence to all prompt elements, including specific food categories and color accents.
- − Overall design is a bit conservative or 'stock-photo' in style.
Seedream 4.0
- + Bold, clear heading text.
- + Modern, high-energy photo grid arrangement.
- − Fails to provide actual menu content like item names or prices.
- − The layout is just a collage of headers and photos rather than a functional menu design.
- − Large amount of wasted white space in the center of the design.
Verdict: GPT Image 1.5 is the clear winner as it produced a fully realized, professional menu design with legible text, item descriptions, and a logical layout. Seedream 4.0 created an abstract collage that lacks the functional elements of a menu, such as item lists and prices, making it unusable for the requested purpose.
Geometric Composition
Text-to-Image“A glass cube on a wooden table. Inside the cube is a small blue sphere. On top of the cube sits a red book. A green plant is behind the cube, partially visible through the glass. Soft window light from the left.”
AI Judge Analysis
GPT Image 1.5
- + Excellent photographic realism and sharpness
- + Exceptional handling of glass reflections and physical thickness
- + Perfect adherence to all spatial instructions and lighting
- − The sphere is quite large relative to the prompt 'small blue sphere'
Seedream 4.0
- + Successfully placed all requested elements in the scene
- + Accurately depicted a 'small' blue sphere as requested
- + Strong sense of cinematic natural lighting
- − The glass cube is missing its back-left vertical edge, making the geometry incoherent
- − Significant artifacting where the plant is visible through the glass
Verdict: Both models followed the prompt instructions perfectly in terms of object placement and lighting direction. GPT Image 1.5 is the clear winner due to its superior image quality and physical coherence; Seedream 4.0 suffers from broken geometry on the glass cube and messy rendering of the background plant through the glass.
Man and Car in California
Editing“Make a photo of the man driving the car down the California coastline”
AI Judge Analysis
GPT Image 1.5
- + Excellent preservation of the man's facial features and hair texture.
- + High-quality rendering of the car's interior details.
- + The lighting on the man matches the beach environment well.
- − The perspective of the car feels slightly disjointed from the road.
- − The man is positioned too far back in the seat relative to the steering wheel.
Seedream 4.0
- + Perfectly preserves the specific car model (Rolls-Royce) and its details from the source image.
- + Dynamic and realistic composition with a clear sense of motion.
- + Accurately represents the man's clothing and hairstyle in the new context.
- − The man's facial details are slightly blurred compared to the source image.
- − Minor jitter in the wheel spokes due to the motion blur effect.
Verdict: Both models performed exceptionally well, but Seedream 4.0 is the winner for its superior composition and preservation of the car's physical presence. While GPT Image 1.5 kept the man's face sharper, the overall scene in Seedream 4.0 feels much more natural and cohesive as a single photograph.
Bald man challenge
Editing“Give the person a full, thick head of natural hair with realistic texture, density, and a natural hairline. Preserve facial features and lighting.”
AI Judge Analysis
GPT Image 1.5
- + Natural hair texture that matches the beard style
- + Seamless integration of hair with the original facial features
- + Excellent preservation of background and lighting
- − Slightly altered the shape of the glasses frames
Seedream 4.0
- + Followed the instruction for 'thick' hair very literally
- + Maintained the original background and lighting well
- − Hair looks like a wig with an unnatural, stiff texture
- − The hairline is too low and lacks realistic transition
- − The vertical volume of the hair feels out of proportion for a natural head of hair
Verdict: GPT Image 1.5 provides a much more convincing and realistic edit, creating hair that matches the character's facial hair and sits naturally on the head. Seedream 4.0 added a very large volume of hair that lacks realistic texture and looks more like a toupee or a wig, failing the 'natural' requirement of the prompt.
Night Sky Transformation
Editing“Change the scene to night: a deep, dark sky with subtle, glistening stars visible behind the mountain.”
AI Judge Analysis
GPT Image 1.5
- + Excellent transformation of the atmosphere to a true night scene
- + Impressive starry sky with realistic density and distribution
- + Accurately darkened the village and landscapes while maintaining city lights
- − The mountain peak is perhaps a bit too dark compared to a moonlit night
Seedream 4.0
- + Maintains the specific architectural details of the original village more accurately
- + Good balance of light on the mountain face
- + Preserves the foreground textures well
- − The sky is not as deep or dark as requested, feeling more like twilight
- − The stars are sparse and less impactful compared to the prompt's 'glistening' request
Verdict: GPT Image 1.5 followed the prompt much more effectively, delivering a deep, dark night sky with a dense field of glistening stars. While Seedream 4.0 preserved more of the source image's sharpness, it failed to fully transition the scene to night, resulting in an unnaturally bright landscape and a relatively empty sky.
Over-the-top cartoon caricature
Editing“Create a caricature of me and my job. Make it exaggerated and humorous, incorporating my profession as a tv show anchor and my love for dogs and hockey.”
AI Judge Analysis
GPT Image 1.5
- + Excellent caricature style with highly exaggerated features that suit the request.
- + Rich, vibrant background incorporating all requested themes (news desk, dogs, hockey game).
- + High-quality rendering with no major anatomical issues or artifacts.
- − Changes the character's clothing and eye color, losing some of the source identity.
- − Text is slightly generic compared to the visual density.
Seedream 4.0
- + Maintains the subject's original outfit and features more accurately than Model A.
- + Preserves the background elements of the source image while overlaying the news desk.
- + Cleverly combines the hockey and anchor themes through the jersey and stick placement.
- − The caricature deformation on the face has some artifacts around the mouth and eyes.
- − The composition feels a bit cluttered with the phone, microphone, and desk overlapping awkwardly.
Verdict: GPT Image 1.5 delivers a much more polished and 'exaggerated' caricature as requested, featuring a creative integration of a dog in a hockey helmet and a full news studio background. Seedream 4.0 does a better job of preserving the specific person's likeness and clothing from the source image but suffers from messy visual artifacts and a less appealing overall composition.
Adorable Baby Animals in Sunny Meadow
Text-to-Image“Hyper-photorealistic scene of fluffy baby animals—a golden retriever puppy, tabby kitten, baby bunny, and red fox kit—with big expressive eyes and ultra-detailed soft fur, playfully chasing butterflies and tumbling together in a lush wildflower meadow, warm golden sunrise light with god rays and dew sparkles, joyful wholesome vibe, 8K masterpiece.”
AI Judge Analysis
GPT Image 1.5
- + Excellent fur texture and fine detail on all four animals.
- + Expressive facial features that perfectly match the 'joyful' vibe.
- + Stronger adherence to the 'tumbling together' request with close, intertwined positioning.
- − The kitten has an anatomically strange third hind leg/paw visible at the bottom.
- − The fox's front paw has a slightly merged, undefined look.
Seedream 4.0
- + Dynamic composition that captures the 'chasing' aspect better by showing the animals in motion.
- + Beautiful bokeh and lighting effects with well-distributed dew sparkles.
- + The animals are clearly distinguishable and appropriately sized relative to one another.
- − The kitten has a very small, somewhat distorted front paw while reaching up.
- − The fox kit's red fur is slightly over-saturated, bordering on unrealistic compared to the other animals.
Verdict: Both models followed the complex prompt extremely well, including all four specific animals and the difficult lighting conditions. GPT Image 1.5 wins on fur texture and emotional expression, creating a very heartwarming close-up, though it suffers from a significant anatomical limb error on the kitten. Seedream 4.0 is preferred for its superior composition and action, capturing a more believable sense of 'playing and chasing' in a meadow without major anatomical failures.
Victorian Greenhouse Oasis
Text-to-Image“Hyper-photorealistic interior of a lush Victorian glass greenhouse filled with exotic tropical plants, vibrant blooming orchids, tall ferns, colorful butterflies in flight, sunlight filtering through ornate glass roof creating realistic caustics and dew on leaves, intricate iron framework visible, misty atmosphere, 8K masterpiece.”
AI Judge Analysis
GPT Image 1.5
- + Exceptional level of detail in the iron framework and glass textures.
- + Beautiful rendering of dew drops and sunlight filtration through the atmosphere.
- + Lush, dense composition that feels much more like a 'masterpiece' with high complexity.
- − The butterflies appear somewhat flat and lack the realistic motion or lighting integration seen in the rest of the scene.
Seedream 4.0
- + Good use of depth of field with a clear focal point on the foreground orchids.
- + Effective misty atmosphere that creates a soft, ethereal mood.
- − The iron structure is much simpler and lacks the 'intricate' Victorian detail requested.
- − The lighting and caustics are less defined compared to the other image.
- − Less plant variety and density visible in the mid-ground.
Verdict: GPT Image 1.5 is the clear winner as it more successfully captures the 'intricate iron framework' and '8K masterpiece' aspects of the prompt with stunning detail in the water droplets and architecture. Seedream 4.0 provides a pleasant image with good depth of field, but it feels sparse and less technically impressive in its rendering of the Victorian greenhouse structure.
Heroic Super Hero Portrait
Text-to-Image“Hyper-photorealistic full-body portrait of a female superhero standing triumphantly on a New York skyscraper rooftop at golden sunset, wearing a classic modest superhero costume with flowing cape, chest emblem, gloves, and boots in red and blue colors, practical design, short hair, strong determined heroic expression looking into the distance, powerful confident stance with hands on hips and cape billowing dramatically in the wind, detailed urban cityscape background, warm natural sunlight with sharp shadows and fabric highlights, ultra-sharp textures on suit, hair, and concrete, 8K masterpiece, empowering family-friendly style.”
AI Judge Analysis
GPT Image 1.5
- + Excellent skin texture and fabric detail realism.
- + Dynamic and large-scale billowing of the cape.
- + Consistent and vibrant golden hour lighting across the character and background.
- − The skirt design makes the costume less 'practical' compared to the other model.
- − Lighting on the character's face is a bit flat despite the strong backlight.
Seedream 4.0
- + More practical and modern superhero suit design with pants.
- + Superior cinematic lighting with a natural rim light effect on the hair and silhouette.
- + More expressive and atmospheric composition that feels less like a studio setup.
- − The cape physics are slightly less dramatic than requested.
- − Missing the requested gloves.
Verdict: GPT Image 1.5 offers a very high-resolution and detailed look with impressive textures, though the costume interpretation is a bit more traditional/skirted. Seedream 4.0 captures a more cinematic and modern aesthetic with a more practical suit design and superior lighting, though it missed the specific detail of gloves. Overall, Seedream 4.0's composition and lighting feel more authentic to a professional film still.
Neutral Expression to Genuine Smile
Editing{
"action": "image_edit",
"reference": "uploaded neutral portrait",
"change": "Warm genuine Duchenne smile: lips curved up, slight natural teeth, soft eye crinkles, subtle cheek raise",
"details": "Realistic smiling skin (dimples if present, soft cheek shadows), slightly brighter eyes; keep exact eye shape/color/iris",
"preserve_exact": "Face identity/structure, eyes/nose/lips/eyebrows, hair, skin texture/pores/freckles, makeup, clothing, head pose, background, lighting, shadows, framing",
"no_changes": "No face shape change, no new features, no gaze shift, no hair/clothing/lighting/background edits",
"style": "Ultra-photorealistic 8K portrait, sharp face focus, natural soft lighting, realistic skin glow"
}
AI Judge Analysis
GPT Image 1.5
- + Perfect preservation of original skin texture and freckles
- + Maintains exact facial identity and eye shape
- + Exceptional resolution and sharpness
Seedream 4.0
- + Accurately depicts the Duchenne smile with eye crinkles and eye narrowing
- + Good preservation of hair and background
- + Warmer, natural color palette
- − Slightly softens original skin texture/pores
- − The smile is a bit wide, slightly altering the original jawline shape
Verdict: GPT Image 1.5 is the winner because it achieves perfect preservation of the original image's texture and identity; every freckle and pore remains exactly as it was in the source while adding a very natural smile. Seedream 4.0 performs well on the emotional expression of the smile (especially the eyes), but it slightly smooths the skin and loses some of the specific fine-grain detail from the original.
Studio Ghibli Anime Style
Editing“Transform this photo into a Studio Ghibli–inspired illustration. Use soft pastel colors, hand-painted textures, gentle lighting, dreamy backgrounds, and a warm, nostalgic mood”
AI Judge Analysis
GPT Image 1.5
- + Excellent preservation of the original subjects' clothing patterns and poses
- + Captures the warm, nostalgic, and dreamy lighting requested in the prompt
- + Maintains the depth of field from the source image with the foreground girl blurred
- − Art style leans more towards generic shojo anime than distinct Studio Ghibli aesthetics
- − Faces are a bit too modernized for a nostalgic Ghibli look
Seedream 4.0
- + Perfectly captures the Studio Ghibli hand-painted watercolor aesthetic
- + Highly accurate character facial expressions that match the source while adopting the target style
- + Exceptional execution of hand-painted textures and soft pastel colors
- − The dreamy background replaces the urban street with abstract colors, losing some environmental context
- − The foreground character's blur is removed, making the composition flatter than the original
Verdict: Both models successfully interpreted the meme prompt and maintained the core composition. GPT Image 1.5 did a better job of preserving the source image's layout and clothing details, but Seedream 4.0 far exceeded it in terms of artistic style accuracy, delivering a near-perfect Ghibli watercolor aesthetic that fits the 'hand-painted' and 'nostalgic' requirements precisely.
Intricate Floral Mandala
Text-to-Image“Perfectly symmetrical mandala made entirely of real flowers, petals, leaves, fruits, and seeds in vibrant natural colors, intricate layered patterns with radial symmetry, top-down view on a soft neutral background, hyper-detailed organic textures and subtle shadows, photorealistic, 8K masterpiece.”
AI Judge Analysis
GPT Image 1.5
- + Excellent adherence to the 'perfectly symmetrical' requirement with precise radial patterns.
- + High clarity and separation of distinct natural elements like seeds, berries, and petals.
- + Exceptional top-down composition that mimics traditional mandala art.
- − The elements look slightly more like plastic or clay than organic matter in some areas.
- − Lighting is very flat, lacking the atmospheric depth requested by 'subtle shadows'.
Seedream 4.0
- + Superior organic textures and realistic lighting with soft, natural shadows.
- + Better variety in the 'fruits' category, including identifiable pears, oranges, and berries.
- + Elements feel like real physical objects resting on a surface.
- − Failed the 'perfectly symmetrical' requirement, with many elements placed haphazardly.
- − Center of the mandala is muddy and lacks the intricate pattern defined in the prompt.
- − The 'top-down' view is slightly angled rather than being a true flat-lay.
Verdict: GPT Image 1.5 is the clear winner because it successfully delivers on the primary structural requirement of a 'perfectly symmetrical mandala.' While Seedream 4.0 has more convincing organic textures and realistic lighting, its lack of symmetry and chaotic arrangement makes it fail the core geometric definition of a mandala.
Golden Hour Stroll
Editing“Add dynamic motion to this photo: make hair blow in the wind, add leaves flying, energetic and lively feel.”
AI Judge Analysis
GPT Image 1.5
- + Successfully added a large volume of flying leaves for a high-energy feel.
- + The hair is dramatically windswept, effectively conveying motion.
- + Maintains high image clarity and facial details from the original.
- − The leaves appear pasted on top and don't always interact realistically with the depth of the scene.
- − Slightly altered the woman's facial features compared to the source image.
Seedream 4.0
- + Excellent preservation of the woman's original face and expression.
- + The wind effect on the hair is realistic and well-integrated.
- + Leaves are placed with a better sense of depth and motion blur.
- − Significantly lower resolution and overall sharpness compared to the source and Model A.
- − Fewer flying leaves makes the scene feel less 'energetic' than requested.
Verdict: Both models followed the instructions well, but GPT Image 1.5 produced a much sharper, high-resolution result with a more intense 'energetic' feel due to the quantity of flying leaves. While Seedream 4.0 did a better job of preserving the subject's original facial features and creating a more natural sense of depth, the significant loss in image quality makes it the less desirable output.
Vintage Cafe Logo
Text-to-Image“Vintage minimalist restaurant logo for "Caffè Florian", retro cloche dome with steam and "Est. 1720" banner, classic typography, warm brown and cream tones, subtle texture on light background, vector emblem style.”
AI Judge Analysis
GPT Image 1.5
- + Excellent classic typography with a custom feel
- + Included the grave accent on 'Caffè' correctly
- + Good use of texture and shading on the cloche
- − Ignored the request for a light background
- − Banner integration is slightly clunky compared to the text
Seedream 4.0
- + Followed all prompt instructions including the light textured background
- + Clean and balanced vector-style composition
- + Legible and well-spaced typography
- − Used an acute accent (é) instead of the requested grave accent (è)
- − The steam lines are somewhat generic
Verdict: Both models followed the core elements of the prompt well, including the specific date and cloche imagery. Seedream 4.0 followed the background instructions more accurately and feels more like a finished logo emblem, whereas GPT Image 1.5 ignored the background requirement but produced superior, more authentic typography.
Apollo 11: Journey to Tranquility
Text-to-Image“Create a clean, modern vector infographic poster about the Apollo 11 mission. NASA-inspired palette (navy, white, muted red, light gray). Flat-vector style, crisp lines, consistent iconography, subtle gradients only. Steps (stop at landing): 1. Launch (Saturn Vicon) 2. Earth Orbit (Earth + orbit ring icon) 3. Translunar (trajectory arc icon) 4. Lunar Orbit (Moon + orbit ring icon) 5. Descent (lunar module descending icon) 6. Landing (lunar module on the surface icon) Small supporting elements (minimal text): • Crew strip: three silhouette icons with only last names: Armstrong, Aldrin, Collins. • Landing site marker: Moon pin labeled "Tranquility" only. Layout constraints: generous margins, large readable labels, clean background with subtle stars. Vector-only, print-poster look, high resolution.”
AI Judge Analysis
GPT Image 1.5
- + Perfect adherence to all six requested steps with appropriate icons.
- + Exceptional text rendering for all labels including crew names.
- + Highly professional vector aesthetic with a consistent and clean layout.
- − One minor text cut-off at the very top of the image.
Seedream 4.0
- + Accurately followed the color palette and basic steps.
- + Clean, readable typography for the main title and numbered steps.
- + Included the requested 'Tranquility' location marker.
- − Confusing visual layout where the Lunar Module is used for both 'Descent' (5) and 'Landing' (6) without a clear icon change.
- − Icons are significantly less detailed and polished compared to the other model.
- − Text for 'Translunar' is slightly cramped/misaligned.
Verdict: GPT Image 1.5 is the clear winner as it provides a professional, well-balanced infographic with high-quality icons for every single requested step. Seedream 4.0 followed the instructions well but suffered from simpler graphics and a layout that didn't visually distinguish the final stages as effectively as GPT Image 1.5.
GPT Image 1.5
OpenAI's state-of-the-art image generation model with better instruction following and adherence to prompts
Seedream 4.0
ByteDance's image generation model with integrated text-to-image and image editing capabilities in a unified architecture, supporting up to 4K resolution