GPT Image 1.5 vs Seedream 4.5
Head-to-head across 18 challenges
GPT Image 1.5
57.1%
win rate
Ties
7.1%
Seedream 4.5
35.7%
win rate
Challenge Results
Magic Burger Explosion: Fiery Photorealism Challenge
Text-to-Image“Ad for 'Magic Burger'. Dynamic, exploded burger with all components (bun, patty, cheese, lettuce, tomato, sauce) suspended in mid-air. Emphasize photorealistic detail and a sense of motion. Dark, fiery background with glowing embers. Integrate text: 'MAGIC BURGER' as a prominent title, 'LIMITED TIME ONLY' as a secondary message, and '€6.99' in a starburst, all rendered with a fiery, glowing effect.”
AI Judge Analysis
GPT Image 1.5
- + Excellent photorealistic texture on the meat and bun
- + Very effective use of the fiery, glowing effect on all text elements
- + Dynamic composition with embers and sauce splatters that enhance the action
- − The 'exploded' effect is a bit vertical and static compared to Model B
- − The bottom bun remains at the very bottom of the frame rather than being fully suspended
Seedream 4.5
- + Stronger sense of motion with diagonal arrangement and motion blur trails
- + Cleaner text layout and typography
- + Creative interpretation of the cheese stretching between components
- − The textures look slightly more digital/artificial than Model A
- − The lettuce and tomato look less realistic and more like plastic props
Verdict: GPT Image 1.5 wins on photographic realism and the integration of the fiery theme into the burger's textures, making the food look more appetizing. While Seedream 4.5 has a more dynamic 'exploded' composition and cleaner text, it lacks the gritty, high-detail realism found in GPT Image 1.5's rendering of the patty and bun.
Outfit Transfer Challenge
Editing“Use Image 1 as the base person. Dress them in the exact elaborate outfit from Image 2 (including all layers, accessories, jewelry, and shoes). Carefully adapt the clothing to the body shape and pose in Image 1 while maintaining realistic fabric behavior, correct proportions, and perfect lighting/shadow matching. Keep the person’s exact face, hair, and background completely unchanged.”
AI Judge Analysis
GPT Image 1.5
- + Excellent preservation of the source subject's identity, facial structure, and hair.
- + Perfect execution of clothing transfer, including the pea coat, scarf, and watch.
- + Maintains the original lighting and background consistency correctly.
Seedream 4.5
- + Successfully transfers the clothing items including accessories like the ring and watch.
- − Fails to keep the person's identity unchanged by blending the faces of the two source images.
- − The hair was changed to match the person in Image 2 instead of Image 1.
- − Lower resolution and poorer integration of the skin texture on the chest.
Verdict: GPT Image 1.5 is the clear winner as it followed all instructions perfectly, preserving the identity of the person in Image 1 while flawlessly adding the clothing from Image 2. Seedream 4.5 failed the primary 'preservation' constraint by creating a hybrid face of both individuals and changing the subject's hairstyle.
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
- + Successfully merges the subject and car from the source images into the requested new environment.
- + Preserves the subject's facial features and distinctive hairstyle very accurately.
- + Excellent lighting and color matching between the car interior, the driver, and the coastal background.
- − The steering wheel is positioned incorrectly, appearing to grow out of the middle of the dashboard rather than being in front of the driver.
- − The subject is not actually holding the steering wheel correctly.
Seedream 4.5
- + Excellent full-body preservation of the subject's outfit, including the specific coat, scarf, pants, and boots.
- + High degree of source car preservation, including the interior details and door shape.
- + Composition shows a clearer view of both the car and the coastline.
- − The car door is wide open while driving, which is a significant logical error.
- − The subject's face is slightly altered and less accurate to the source photo compared to Model A.
Verdict: Both models do an impressive job of combining elements from two disparate source images into a single scene. GPT Image 1.5 achieves a more natural lighting and facial resemblance, but fails on the interior ergonomics with a misplaced steering wheel. Seedream 4.5 captures the most detail from the source clothing and car, but suffers from the nonsensical logic of the driver's door being fully open while the car is in motion.
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
- + Perfect adherence to the prompt's spatial instructions.
- + High-quality textures on the red book and wooden table.
- + Realistic reflections and refractions through the glass cube.
- − The 'small blue sphere' is relatively large compared to the cube.
Seedream 4.5
- + Captures the soft window light effect with high contrast and warmth.
- + Authentic glass refraction for the plant in the background.
- − Serious geometric errors where the blue sphere is merging into/outside the glass wall.
- − The cube shape is distorted and looks like a solid block rather than a hollow container.
- − The plant is very blurry and barely recognizable.
Verdict: GPT Image 1.5 successfully followed all spatial instructions, placing the sphere clearly inside the cube and the plant behind it. Seedream 4.5 struggled with spatial reasoning, resulting in a blue sphere that appears to be clipped through the glass wall and a distorted cube structure.
Pose & Character Mashup
Editing“Use Image 1 as the exact pose reference and Image 2 as the character reference. Recreate the person/character from Image 2 in the exact dynamic pose and body position from Image 1. Keep the exact face, hair, clothing style/details, and expression from Image 2. Match the lighting and environment of Image 1. The final image must show the character from Image 2 performing the precise action/pose from Image 1 with perfect anatomy and natural integration.”
AI Judge Analysis
GPT Image 1.5
- + Near-perfect adherence to the scarf pattern and facial features from Image 2.
- + Accurately recreates the lighting and yellow background from Image 1.
- + High resolution with clean textures on the clothing.
- − The head angle is upright instead of tilted to match the pose in Image 1.
- − One foot is drawn with six toes.
Seedream 4.5
- + Captures the tilted head angle and dynamic leaning torso of Image 1 much better than Model A.
- + Good preservation of the character's accessory details like the scarf and sunglasses.
- + Skin tone and facial hair are consistent with the source character.
- − Major anatomical error where the character's left leg appears to grow out of their stomach.
- − Small artifacts around the hands and shoe/foot areas.
Verdict: GPT Image 1.5 succeeds in preserving the exact identity and details of the character from Image 2, but it fails to replicate the specific head tilt of the pose in Image 1. Seedream 4.5 manages to capture the dynamic lean and head orientation of the pose much more accurately, but it suffers from a significant anatomical failure regarding the leg structure. GPT Image 1.5 is the winner for its superior image quality and closer adherence to the scarf's visual pattern, despite the minor toe count error.
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 photorealistic texture on the jacket, skin, and bicycle.
- + Successfully incorporates the 'imperfect framing' prompt with a large car obscuring part of the scene.
- + Highly detailed and realistic rendering of a wet pavement environment.
- − The 'motion blur' on the car is quite minimal, appearing more as a static object than a moving one.
Seedream 4.5
- + Strong execution of motion blur from passing cars in the background.
- + Good adherence to the 'shallow depth of field' and 'candid' look.
- + Natural skin textures and realistic expressions.
- − The bicycle mechanics are nonsensical, with a wrench floating near a chain that isn't connected to a sprocket correctly.
- − Composition is a bit more 'posed' rather than 'candid' as the subject is looking directly at the camera.
Verdict: GPT Image 1.5 is the winner due to its superior technical accuracy; while Seedream 4.5 captures the motion blur of cars better, it fails significantly on the details of the bicycle repair. GPT Image 1.5 feels like a genuine, high-quality street photograph with realistic textures and a believable environment.
The Reversed Rodeo
Text-to-Image“Horse riding astronaut in space — horse on top, not vice versa. Surreal, highly detailed, cinematic.”
AI Judge Analysis
GPT Image 1.5
- + Excellent cinematic lighting and complex background composition
- + High level of texture detail on the astronaut suit and horse's fur
- + Coherent environmental details like lunar dust and various celestial bodies
- − Failed the negative constraint entirely by placing the astronaut on top of the horse
Seedream 4.5
- + Successfully followed the difficult spatial constraint of placing the horse on top of the astronaut
- + Vibrant and aesthetic color palette in the nebula
- + Expressive horse anatomy and dynamic posing
- − Anatomical issues where the astronaut's right leg appears to blend into or come out of the horse's flank
- − The horse's bit and bridle are poorly integrated with its mouth
Verdict: GPT Image 1.5 produced a much higher quality image in terms of detail and realism, but failed to follow the specific 'horse on top' instruction. Seedream 4.5 successfully interpreted the surreal prompt correctly despite having some anatomical merging issues, making it the better choice for the specific request.
The Capybara Taxi Driver
Text-to-Image“Photorealistic scene inside a yellow New York taxi at night. A capybara is driving, wearing a yellow taxi driver cap and a dark jacket. It has a calm, professional expression and both front paws on the steering wheel. In the back seat sits a human businesswoman in a coat, looking at her phone with a completely normal, bored expression (as if this is just another normal ride). Through the windows you can see the streets of Manhattan at night with blurred lights. Realistic taxi interior, photorealistic, detailed fur and fabric, 35mm lens, night lighting with reflections, shallow depth of field.”
AI Judge Analysis
GPT Image 1.5
- + Excellent texture on the capybara's fur and the jacket fabric.
- + Cinematic lighting and depth of field that creates a gritty, realistic NYC atmosphere.
- + The passenger has a perfect bored expression as requested.
- − The capybara's paws look slightly mutated or claw-like.
- − The steering wheel is oddly shaped and too small in scale.
Seedream 4.5
- + Clean composition with a very clear view of both subjects.
- + The capybara's expression is very calm and professional.
- + The cap and taxi lights are bright and well-integrated.
- − The fur texture looks slightly smoothed and less realistic than Model A.
- − The passenger's face looks slightly blurred or lower resolution compared to the driver.
Verdict: GPT Image 1.5 wins due to its superior photorealism and cinematic atmosphere, which better captures the 'moody' night-time NYC setting. While Seedream 4.5 offers a clearer composition and better anatomy on the paws, it has a flatter, more digital appearance compared to the highly detailed textures in GPT Image 1.5.
Chalkboard Menu
Text-to-Image“Handwritten-style chalkboard menu in a cozy café, all text rendered in the exact same realistic chalk handwriting style with natural variations in letter size, slight slant, and chalk texture — no printed or digital fonts anywhere on the board. Title at the top in elegant cursive chalk handwriting: ‘TODAY’S SPECIALS – APRIL 30, 2026’. Below it, three menu items also in the same handwritten chalk style: ‘Truffle Mushroom Risotto – $24’, ‘Grilled Octopus with Lemon & Herbs – $28’, ‘Brown Butter Chocolate Chip Cookies – $9’. At the very bottom, smaller text in the identical handwritten chalk style (slightly smaller but still clearly legible with the same handwriting characteristics): ‘All items made fresh daily • Ask about our gluten-free options’. Warm ambient café lighting, visible chalk dust and smudges, realistic handwriting imperfections, no clean printed text anywhere.”
AI Judge Analysis
GPT Image 1.5
- + Excellent chalk texture with realistic smudging and dusting on the board.
- + Flawless text rendering with consistent handwriting style throughout.
- + Accurately completed the partial prompt 'Brown But...' with a logical 'Chocolate Chip Cookies' addition.
Seedream 4.5
- + Great environmental context showing a cozy café interior.
- + Correctly followed the menu list and prices with good readability.
- + The handwriting looks authentic and utilizes the vertical space of the board well.
- − Repeated the first menu item 'Truffle Mushroom Risotto - $24' twice in a row.
- − The text style at the bottom looks slightly more like a digital font than natural handwriting compared to the top.
Verdict: GPT Image 1.5 followed the prompt instructions perfectly, specifically excelling at the 'chalk texture' and providing more consistent, human-like handwriting across the entire board. Seedream 4.5 had a redundancy error where it repeated the first menu item, although it provided a better environmental background.
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 menu items and descriptions.
- + Logical and highly professional layout that mimics a real-world menu.
- + High-quality, appetizing food photography that perfectly aligns with the text items.
Seedream 4.5
- + Clean, minimalist aesthetic with bold colored borders.
- + Follows the general section structure requested in the prompt.
- − Severely garbled and repetitive text (e.g., 'Appetizters', 'Festaurant').
- − Excessive whitespace and lack of detailed item descriptions.
- − Menu prices are unrealistically high ($79 for an appetizer).
Verdict: GPT Image 1.5 is the clear winner as it produces a fully functional, professional-grade menu with perfect text legibility and logical itemization. Seedream 4.5 follows the minimalist prompt but fails significantly on text rendering, producing nonsensical words and repetitive placeholders that make the design unusable.
The Halloween Invitation
Text-to-Image“Vintage gothic Halloween party invitation. Dark parchment poster, spooky border with webs and thorns, central glowing jack-o-lantern, bats, twisted trees, moody night sky. Add elegant gothic title text saying "Halloween Party Invitation", a small scroll banner saying "You are invited to a night of frights", and event details at the bottom: Date: 30.10.2026 Time: 7pm Location: The Arches, NYC Spooky but polished, cinematic lighting, square format.”
AI Judge Analysis
GPT Image 1.5
- + Perfect text rendering for all requested fields including specific dates and addresses.
- + Highly cohesive aesthetic where the gothic border, parchment texture, and illustration blend seamlessly.
- + Superior cinematic lighting and atmosphere that feels like a vintage printed poster.
- − The background elements like the trees and castle are slightly less sharp than the central pumpkin.
Seedream 4.5
- + Clean, vibrant colors with a high-contrast cinematic glow around the pumpkin.
- + Accurate text rendering for the primary titles and event details.
- + Distinct parchment-style frame that makes the central image pop.
- − The border features barbed wire instead of the requested thorns.
- − The composition feels slightly disjointed, with the text overlaying the moon and trees in a less integrated way.
- − The placement of the location and date text at the bottom is less balanced than Model A.
Verdict: GPT Image 1.5 is the winner because it captures the 'vintage gothic' aesthetic perfectly, integrating the text and illustration into a single cohesive piece of art. Seedream 4.5 produces a high-quality image, but the use of barbed wire instead of thorns and the less elegant layout of the event details makes it feel less like a polished invitation.
Bald man challenge
Image 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 wavy hair texture that matches the beard style
- + Excellent preservation of the original facial features and expression
- + Seamless integration of the hairline and sideburns
Seedream 4.5
- + Matches the requested 'thick head of hair' prompt
- + Good preservation of the background and clothing
- − The hairline looks slightly artificial and overly rounded
- − Visible distortion/smudging on the right eye and eyelid compared to the source
- − The hair texture appears a bit flat and painted-on near the forehead
Verdict: GPT Image 1.5 is the winner because it provides a much more natural-looking hair texture and hairline that perfectly complements the subject's existing beard. Seedream 4.5 slightly alters the person's facial features, particularly around the eyes, and the resulting hairline looks less realistic.
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
- + Captures all requested elements including news setting, multiple dogs, and hockey in a vibrant TV studio environment.
- + Strong artistic caricature style with exaggerated features and high energy.
- + Excellent text rendering and graphic design for the 'Breaking News' overlay.
- − Completely changes the subject's outfit and background, losing the source image context.
- − The character's hands and the objects they are holding have some anatomical and structural warping.
Seedream 4.5
- + Preserves the subject's original denim outfit and the living room background from the source image while adding the desk.
- + The hockey gear and dog are rendered with high clarity and realistic textures.
- + The caricature head-to-body ratio is well-executed for a 'bobblehead' style.
- − The microphone setup is physically disconnected from the desk, appearing to float.
- − Less 'humorous' and 'exaggerated' in its overall composition compared to Model A.
- − The desk and background integration feels slightly mismatched in terms of lighting.
Verdict: GPT Image 1.5 creates a much more cohesive and imaginative scene that feels like a professional caricature, successfully incorporating the hockey theme into the background action and the dogs. Seedream 4.5 does a better job of preserving the source image's clothing and background but suffers from technical glitches like a floating microphone and a less dynamic interpretation of the prompt.
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 unavailable for this challenge.
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 unavailable for this challenge.
Golden Hour Stroll
Image 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
- + Excellent preservation of the subject's face and clothing.
- + The hair-blowing effect is realistically integrated with the original pose.
- + High density of falling leaves creates a strong sense of wind.
Seedream 4.5
- + Successfully captures a more 'energetic' feel by slightly altering the pose to a jog.
- + The leaves have motion blur which enhances the sense of dynamic movement.
- + Good preservation of the background and surroundings.
- − The leash now appears to be floating unconnected to the dog's collar.
- − The subject's face has changed significantly from the source image.
- − The hand holding the leash has anatomical issues/distortion.
Verdict: GPT Image 1.5 is the winner because it successfully applied the dynamic edits while maintaining perfect consistency with the source image's subject and details. Seedream 4.5 created a more energetic composition, but at the cost of changing the woman's face and introducing a noticeable error where the leash no longer connects to the dog.
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 typography with correct Italian accent usage.
- + Rich vintage texture and sophisticated vector shading.
- + Strong central composition with a classic emblem feel.
- − Ignored the request for a light background, providing a black background instead.
- − Slightly less 'minimalist' than Model B due to detailed shading.
Seedream 4.5
- + Perfect adherence to the light background and minimalist vector style.
- + Very clean typography and iconography.
- + Clear execution of the banner and steam elements.
- − The 'f' in Florian is slightly disconnected or stylized in a Way that looks like a gap.
- − Shading on the cloche is very basic compared to the artistic depth of Model A.
Verdict: Both models followed the complex text requirements perfectly. Seedream 4.5 adhered better to the background and minimalism constraints of the prompt, while GPT Image 1.5 produced a much more visually compelling and textured piece of art that failed the 'light background' instruction. Seedream 4.5 is the winner for better prompt adherence regarding color scheme and style.
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
- + Excellent adherence to all step-by-step landing icons
- + Strong typography and legible text for all labels
- + Dynamic and engaging vertical composition
- − The 'Launch' section is slightly cluttered by the overlapping Saturn V rocket
- − Some icons are slightly more illustrative than strictly 'flat vector'
Seedream 4.5
- + Clean, minimalist flat-vector aesthetic that matches the 'modern infographic' request
- + Perfectly legible text and clear numbering of steps
- + Accurate use of the requested NASA-inspired color palette
- − Step 5 (Descent) is represented by a generic satellite icon instead of a descending lunar module
- − The step 3 trajectory arc is disconnected from the other elements
Verdict: GPT Image 1.5 is the preferred image because it followed the specific iconography instructions for every step of the mission, including the lunar module for both descent and landing. While Seedream 4.5 captures the 'flat vector' style more accurately, it failed on step 5 by depicting a satellite instead of a descending lunar module.
GPT Image 1.5
OpenAI's state-of-the-art image generation model with better instruction following and adherence to prompts
Seedream 4.5
ByteDance's latest image generation model unifying text-to-image and image editing in a single architecture, with improved text rendering and 30-40% faster generation than v4.0