GPT Image 1 Mini OpenAI vs Seedream 4.0 ByteDance

GPT Image 1 Mini

• + Excellent photographic quality with realistic textures on the book and table.
• + The red book is sized appropriately to sit on top of the cube.
• + Strong composition with pleasing depth of field and soft lighting.

• − The blue sphere appears to be floating mid-air or fused with the glass edge, defying gravity.
• − The plant is positioned more to the side than 'behind' the cube, missing the refraction opportunity.

Seedream 4.0

• + Perfectly follows the spatial instruction of the plant being visible 'through' the glass.
• + Realistic light refraction and reflections on the glass and wooden surfaces.
• + The sphere sits naturally on the bottom surface of the cube.

• − The cube is not quite a perfect cube (more of a rectangular prism).
• − The red book is slightly smaller than the top surface of the cube, making the stack look less balanced.

GPT Image 1 Mini

• + Excellent preservation of the man's facial features and specific hairstyle
• + Maintains the exact clothing patterns from the source image
• + High visual quality with realistic lighting and motion blur

• − The car's front-end design is modified from the original source

Seedream 4.0

• + Near-perfect preservation of the source car's design and front grille
• + Accurately places the man in the scene with a great California coastline background

• − The man's facial features and clothing details are significantly simplified compared to the source
• − The scale of the man relative to the car feels slightly small

GPT Image 1 Mini

• + Excellent shallow depth of field and bokeh
• + Highly realistic skin texture and facial lighting
• + Strong cinematic atmosphere with moodier tones

• − Missed the request for motion blur from passing cars
• − The bicycle structure has some anatomical inconsistencies near the rear wheel

Seedream 4.0

• + Perfect adherence to the motion blur from passing cars request
• + Includes realistic tools scattered on the ground
• + Clear reflections and textures on the wet pavement

• − The man's scale relative to the bicycle feels slightly off
• − The lighting is flatter and less cinematic than the competitor

GPT Image 1 Mini

• + Excellent high-frequency texture on the skin and facial hair
• + Highly intricate and realistic metal engraving
• + Sophisticated warm lighting and atmospheric depth

• − Missed the request for specific beads in the hair braids
• − The leather straps are present but less prominent than requested

Seedream 4.0

• + Perfect adherence to the 'beads in hair' and 'cloth underlayer' prompts
• + Dynamic lighting with a visible torch source
• + Distinct facial scars and dirt patterns

• − The sparks have a slight motion blur/artifacting look that feels less natural
• − The armor texture is somewhat softer compared to Model A

GPT Image 1 Mini

• + Excellent adherence to the grid layout request.
• + Very clean typography and professional document formatting.
• + Perfect mapping of food images to their specific section headers.

• − The placeholder lines for menu items are completely empty, making it look like a template rather than a menu.

Seedream 4.0

• + High-quality, appetizing food photography.
• + Bold, readable sans-serif fonts.

• − The layout is chaotic and does not follow a clean grid or professional menu structure.
• − Section headers are placed awkwardly and do not align logically with the images.
• − Text is cut off at the bottom and the overall composition lacks professional utility.

GPT Image 1 Mini

• + Excellent text rendering with clean, legible fonts
• + Highly detailed and photorealistic food textures
• + Perfect adherence to all requested price and text elements

• − The composition is a bit static for an 'exploded' view
• − The background is very dark with minimal environmental detail

Seedream 4.0

• + Dynamic sense of motion with swirling trails and fire
• + Effective integration of the fiery background
• + Creative layout with a more 'exploded' feel

• − Failed to follow the price prompt, showing 5.99 instead of 6.99
• − Coherence issues with two separate burger assemblies appearing in the same shot
• − Lower resolution detail on food items compared to the other model

GPT Image 1 Mini

• + Excellent text legibility and neat alignment.
• + Correctly followed all spelling and date requirements.
• + High-resolution chalkboard texture that looks very realistic.

• − Failed to follow the request for 'elegant cursive' for the title.
• − The text looks more like a digital stencil than natural, varied handwriting.
• − The background is a flat wall, missing the 'cozy café' atmosphere.

Seedream 4.0

• + Successfully rendered the title in elegant cursive as requested.
• + Very realistic 'smudged' chalk texture and varying letter slants that mimic true handwriting.
• + Great atmospheric composition with the café background.

• − Minor text overlap issues on the second menu item.
• − The date '2026' is slightly less clean than Model A.
• − Higher amount of chalk dust artifacts might be distracting for some users.

GPT Image 1 Mini

• + Excellent character consistency, accurately representing the man's face, sunglasses, and scarf from Image 2.
• + High image quality with clean lines, smooth gradients, and sharp details.
• + Strong lighting and background match to Image 1's studio yellow.

• − The pose is significantly simplified and doesn't capture the complex, contorted leg and torso position of Image 1.
• − The foot integration on the red box looks slightly flat.

Seedream 4.0

• + Attempts a more complex pose that is closer to the dynamic lean of Image 1.
• + Successfully incorporates the scarf and clothing style from the character reference.

• − Poor character consistency, blending the face and hair of the woman in Image 1 with the man in Image 2.
• − Significant anatomical distortions, particularly with the facial features and the attachment of the right hand.
• − Visible blurring and artifacts throughout the image.

GPT Image 1 Mini

• + Excellent cinematic lighting and atmosphere
• + High detail in the texture of the spacesuit and horse's mane
• + Strong, balanced composition with a moon element

• − Failed the specific positional prompt constraint of horse on top

Seedream 4.0

• + Vibrant colors and high-contrast space background
• + Dynamic posing of the horse and astronaut
• + Good reflection details in the helmet visor

• − Failed the specific positional prompt constraint of horse on top
• − Minor anatomical artifacts where the reigns intersect with the horse's neck

GPT Image 1 Mini

• + Excellent photographic quality and resolution
• + Preserves the vitiligo pattern on the hands and neck very realistically
• + Good integration of the coat and scarf textures

• − Failed to keep the face unchanged, significantly altering the subject's features
• − Missed accessories like the sunglasses and specific shoes
• − The hair and skin pattern on the face do not match the source image exactly

Seedream 4.0

• + Successfully included the accessories like sunglasses and specific shoes from Image 2
• + Maintained the unique hair pattern and pose much better than the competitor
• + High adherence to the 'exact outfit' instruction

• − Added a mustache that was not in either source image
• − Coherence of the face and skin textures is messy and lacks clarity
• − The lighting on the subject feels disconnected from the background

GPT Image 1 Mini

• + Excellent fur texture and photorealistic lighting.
• + The woman's expression perfectly matches the 'bored' instruction.
• + The composition feels more intimate and cinematic.

• − Only one paw is clearly on the steering wheel.
• − The taxi cap looks slightly more like a peaked hat than a standard taxi driver cap.

Seedream 4.0

• + Captures a wider view of the taxi's yellow exterior.
• + Clearly shows both paws on the steering wheel as requested.
• + The lighting on the capybara's face is very clear.

• − The text on the cap is nonsensical ('TAT').
• − The woman's face has slight distortion/artifacts.
• − The transition between the capybara's neck and the jacket looks a bit awkward.

GPT Image 1 Mini

• + Excellent text legibility and alignment
• + Very clean, polished graphic design suitable for a real invitation
• + Accurate rendering of the thorn and web border within the composition

• − Lighting is a bit flat compared to the requested cinematic style
• − The parchment texture is subtle, bordering on a simple dark background

Seedream 4.0

• + Strong cinematic atmosphere and lighting
• + Dynamic and detailed background with twisted trees and a moody moonlit sky
• + Creative interpretation of the thorns and parchment texture

• − Text rendering is messy with several spelling and character artifacts
• − The scroll banner text is nearly illegible
• − The composition feels slightly cluttered

GPT Image 1 Mini

• + Natural wavy hair texture that blends well with the existing beard
• + Good integration with the temples and ears
• + Realistic messy styling that fits the rugged outdoor scene

• − Slightly alters facial features making the subject look like a younger person
• − Glasses frames became thicker and flatter compared to the source

Seedream 4.0

• + Excellent preservation of the subject's original facial features and bone structure
• + Maintains the exact glasses and clothing details from the source image
• + Correct hairline placement according to the prompt

• − Hair texture looks slightly fuzzy and less defined than the beard
• − The hairstyle shape feels a bit artificial and 'helmet-like' around the edges

GPT Image 1 Mini

• + Excellent typography rendering and placement
• + Very clean, soft 3D aesthetic that matches the 'cartoon' request perfectly
• + Superior composition with a more effective use of negative space

• − The raised diorama base is quite thick and large compared to the subject
• − The textures are a bit matte and lack the 'realistic PBR' shine mentioned in the prompt

Seedream 4.0

• + Better variety of sushi types including nigiri, maki, and gunkan
• + More realistic material properties on the fish and ikura (roe)
• + Detailed bamboo mat texture on the diorama base

• − The text 'JAPAN' has a slight artifact/notch in the letter J
• − Composition feels slightly more cluttered compared to the 'minimal' request

GPT Image 1 Mini

• + Successfully incorporates all requested elements: news anchor, dog, and hockey.
• + Features a classic, hand-drawn caricature style with effective facial exaggeration.
• + Clearly displays the subject's profession with a large 'NEWS' sign and microphone.

• − Completely changes the medium of the image from a photo to a pencil/colored drawing.
• − Lacks the specific facial likeness of the source woman compared to the other model.

Seedream 4.0

• + Maintains a high degree of facial recognition from the original source image.
• + Effectively preserves the original background setting while adding professional props.
• + Creative integration of hockey through a jersey and stick on the news desk.

• − The 'bobblehead' style and facial anatomy, particularly the open mouth, look somewhat uncanny.
• − A smartphone is added which wasn't requested, cluttering the composition.

GPT Image 1 Mini

• + Excellent fur texture and realistic lighting on the animals.
• + Clear, well-defined god rays and dewdrops in the background.
• + Accurate representation of all four requested animals with distinct features.

• − The kitten's front right paw looks slightly anatomicaly awkward/blunt.
• − The butterfly wings are a bit simplified in terms of pattern.

Seedream 4.0

• + Dynamic 'tumbling' and 'chasing' poses that perfectly capture the requested energy.
• + Dazzling use of backlighting and sparkling dew effects that create a magical atmosphere.
• + Beautifully detailed butterfly wing patterns.

• − The fox kit's anatomy is slightly distorted, especially the way its body connects to the legs while rolling.
• − The kitten has an extra toe/double paw appearance on its raised left limb.

GPT Image 1 Mini

• + Excellent preservation of the source facial expressions and character features
• + Strong hand-painted texture resembling colored pencils or pastels
• + Maintains the background architecture of the original photo

• − The style leans more toward generic storybook illustration than Takahashi-style Ghibli art

Seedream 4.0

• + Perfectly captures the Studio Ghibli 'Kaguya' or 'Yesterday' watercolor aesthetic
• + Includes the characteristic blush and simple line work of Ghibli character designs
• + Follows the 'dreamy background' instruction with beautiful watercolor bleeds

• − Loses some of the original facial nuance, specifically the sharp jealous look of the woman on the right
• − The man's face is somewhat simplified and generic

GPT Image 1 Mini

• + Successfully added a large number of flying leaves
• + Hair is clearly blowing in the wind
• + Maintains a high level of sharpness across the whole image

• − Significantly altered the woman's facial features and bone structure
• − Background elements like the bridge and trees changed noticeably
• − The dog's face was completely regenerated and looks different

Seedream 4.0

• + Excellent source preservation, keeping the woman and dog's identity almost identical to the original
• + Successfully added the requested 'windy' hair and flying leaves
• + Maintains the background architecture and foliage accurately

• − Low resolution output compared to the other model
• − The image appears slightly blurry or soft throughout
• − Added a lens flare in the top right that wasn't requested

GPT Image 1 Mini

• + Excellent typography rendering with perfect spelling of 'Caffè' including the accent.
• + Clean, vector-like emblem execution.
• + Strong contrast with high legibility.

• − Ignored the request for a 'light background', providing a black one instead.
• − The banner style is a bit flat compared to the cloche.

Seedream 4.0

• + Followed the background instruction perfectly with a subtle paper texture.
• + Excellent vintage illustration style on the cloche and banner.
• + Strong composition with curved overhead text.

• − Minor spelling error in the name, using an acute accent instead of a grave accent (Caffé instead of Caffè).
• − Slightly less 'minimalist' than Model A due to more detailed shading.

GPT Image 1 Mini

• + Excellent typography with perfect spelling and clear, legible fonts.
• + Highly consistent flat-vector style with a professional NASA-inspired aesthetic.
• + Logical flow and layout that clearly follows the requested steps.

• − The 'Translunar' icon is a bit abstract/confusing compared to the others.
• − The rocket icon for Launch is a generic rocket rather than a specific Saturn V silhouette.

Seedream 4.0

• + Dynamic use of space with a starry background and surface-level plane.
• + Includes the names of the crew members as supporting details.

• − Multiple spelling errors including 'Surfcce' and 'Translunar Orbit' appearing twice in the sequence logic.
• − The vector style is less consistent, with some elements appearing 3D or having awkward shadows.
• − Poor text alignment and overlapping elements make it look cluttered.