GPT Image 2 OpenAI vs Seedream 5.0 Lite ByteDance

GPT Image 2

• + Excellent photographic clarity and high resolution textures.
• + Realistic glass refraction and thickness on the edges of the cube.
• + Strong adherence to the lighting prompt with clear soft shadows.

• − None notable.

Seedream 5.0 Lite

• + Successfully follows all prompt instructions including object placement.
• + Good warm color palette and soft atmosphere.

• − Lower resolution with a slightly soft/blurry appearance overall.
• − The lighting is less directional than requested from the left.
• − The glass cube looks thin and less physically grounded compared to the other model.

GPT Image 2

• + Excellent depiction of Japanese street elements including legible vertical text and specific curb details
• + Realistic imperfections like the messy toolbox and slightly cluttered framing
• + Very natural, non-perfect skin texture on the man's face and neck

• − The motion blur on the car in the background looks more like a double exposure than natural motion blur
• − The red bicycle frame has some physical inconsistencies in its structure

Seedream 5.0 Lite

• + Beautiful atmospheric rendering of rain droplets and pavement ripples
• + Effective use of shallow depth of field and authentic light trails from a passing car
• + Stronger emotional quality and focus on the action of repair

• − The bike chain and derailleur are physically impossible and disconnected in some areas
• − The background shop signage has AI-garbled text
• − The skin texture looks slightly more smoothed/digital than Model A

GPT Image 2

• + Excellent texture work on the weathered plate armor and skin
• + Naturalistic lighting and shallow depth of field which feels more cinematic
• + Includes small, subtle beads woven naturally into the hair braids

• − The 'battle-worn' appearance relies more on dirt than prominent scars
• − The armor details become a bit muddy in the shadow areas

Seedream 5.0 Lite

• + Very literal and clear interpretation of 'hair braided with small beads'
• + High contrast torchlight effect gives a dramatic fantasy feel
• + Detailed renderings of leather straps and cloth textures as requested

• − The skin and hair look slightly smooth and CGI-like compared to the other model
• − The scars appear painted on rather than integrated into the skin texture

GPT Image 2

• + Excellent typography and layout design that looks like a real professional menu.
• + High-quality, vibrant food photography that perfectly follows the requested grid structure.
• + Accurate and legible text rendering for descriptions and prices.

• − The logo 'NOVA' has minor texture artifacts within the letters.
• − Slightly cluttered footer compared to the minimalist request.

Seedream 5.0 Lite

• + Strong minimalist aesthetic with plenty of white space.
• + Consistent bold sans-serif fonts throughout the design.
• + Good adherence to the basic category requirements.

• − The 'grid' is a bit clunky with thick borders that feel dated rather than modern.
• − Lower-quality food images that look more like stock photos than professional food photography.
• − Very sparse layout with less visual interest than the competition.

GPT Image 2

• + Excellent photorealistic texture on the meat patty and vegetables
• + Superior text rendering with a complex fiery/lightning effect
• + Dynamic composition with splashes of sauce that enhance the 'exploded' feel

• − The intensity of the background effects can make the layout feel slightly cluttered

Seedream 5.0 Lite

• + Clean, minimalist composition that makes the burger elements pop
• + Accurate placement of all requested text elements
• + Good lighting on the food items consistent with the fire source

• − The meat patty looks a bit dry and flat compared to the other image
• − The 'starburst' for the price is a simple line-art shape rather than a fully rendered effect

GPT Image 2

• + Excellent typography with perfect spelling of all complex menu items.
• + Authentic chalk texture and natural handwriting style that feels hand-drawn.
• + Superior composition with a realistic cafe background and lighting.

• − The date is slightly more blocky than the requested 'elegant cursive' for the title.

Seedream 5.0 Lite

• + Natural-looking chalk board surface with smudges and green tint.
• + Good adherence to the requested layout and date.

• − Multiple spelling errors including 'Heriss', 'Beliter', and 'optoons'.
• − Text appears slightly flat and lacks the realistic chalk grain seen in the other model.
• − Unnecessary underlines slice through the descenders of the letters.

GPT Image 2

• + Excellent character preservation including face, sunglasses, and the patterned scarf.
• + Highly accurate recreation of the complex pose and body alignment from Image 1.
• + Seamless integration of the character's clothing onto the dynamic body position.

• − The fingers on the raised hand are somewhat distorted.
• − Minor clipping issue where the scarf meets the chest area.

Seedream 5.0 Lite

• + Good facial recognition and transfer of the sunglasses.
• + Successfully matches the lighting and vibrant yellow background of the source environment.
• + Effective use of the specific clothing items from Image 2.

• − The pose is less accurate to Image 1, particularly with the tucked-in leg being too thick and poorly positioned.
• − Anatomical issues in the feet and the hands.
• − The scale of the body feels slightly off compared to the stool.

GPT Image 2

• + Excellent adherence to the counter-intuitive instruction of the horse riding the astronaut
• + Highly detailed textures on the spacesuit and lunar surface
• + Clever use of the saddle and harness to make the surreal concept look structural

• − The astronaut's hands/gloves have an anatomical inconsistency with the number of fingers
• − The background stars are somewhat generic

Seedream 5.0 Lite

• + Beautiful, vibrant nebula colors and cinematic lighting
• + Higher compositional energy with the horse in a gallop pose

• − Failed the negative constraint entirely by placing the astronaut on top
• − The horse's tail has weird ghostly artifacts or double-rendering
• − The earth is very small and lacks detail

GPT Image 2

• + Perfect preservation of the original person's face, skin patterns, and expression.
• + Seamless integration of the coat and scarf textures onto the body.
• + Maintains the exact background and lighting of the source image.

• − Missed the sunglasses from the second image.
• − The hand in the pocket looks slightly blurred compared to the source person.

Seedream 5.0 Lite

• + Included the sunglasses from the second source image.
• + Captures the gold watch and ring accessories accurately.

• − Significantly altered the face and changed the direction of the gaze.
• − Lower visual quality with noticeable artifacts around the jawline and hair.
• − Failed to preserve the exact facial structure and unique skin markings of the base person.

GPT Image 2

• + Excellent photorealism with shallow depth of field focusing on the capybara's fur.
• + Accurate interior lighting and realistic taxi partition.
• + Perfectly captures the 'bored' expression of the passenger in the background.

• − The steering wheel placement is a bit low and cramped relative to the capybara's body.
• − Only one paw is clearly visible clutching the wheel.

Seedream 5.0 Lite

• + Shows the full exterior side of the taxi which adds context to the New York setting.
• + Clearer view of both paws on the steering wheel as requested.
• + Sharp details on the vehicle dashboard and exterior livery.

• − The passenger is sitting in the front passenger seat instead of the back seat.
• − The perspective is slightly awkward, making the capybara look like it is floating or misaligned with the seat.

GPT Image 2

• + Excellent layout for an invitation with all text clearly legible.
• + Highly detailed and intricate gothic border featuring skulls, webs, and thorns.
• + Creative inclusion of 'The Arches' as a literal bridge in the background.

• − The date text is slightly smaller than other elements, though still readable.

Seedream 5.0 Lite

• + Strong contrast between the glowing jack-o-lantern and the dark background.
• + Simple, bold text that stands out against the parchment.
• + Strict adherence to the banner and border elements.

• − The transition between the central scene and the parchment background is harsh and looks like a basic overlay.
• − Text rendering on the banner has slight inconsistencies in letter sizing.

GPT Image 2

• + Excellent PBR material rendering with realistic food textures.
• + Perfect adherence to text overlay and flag icon styling.
• + Highly detailed diorama base with consistent isometric perspective.

• − Includes slightly more garnish and objects than the requested 'minimal' amount.

Seedream 5.0 Lite

• + Successfully creates a clean, 3D cartoon style with soft textures.
• + Adheres strictly to the 'minimal garnish' and simple base request.

• − Text and icon are placed to the side rather than the requested top-center.
• − Shadows and lighting are slightly less sophisticated compared to the first model.

GPT Image 2

• + Excellent fur texture and realistic animal proportions.
• + Superior lighting with natural god rays and detailed dew sparkles on the flowers.
• + Strong dynamic composition that feels like a captured moment of movement.

• − The kitten's paw reaching up has a slightly confused anatomical structure.
• − The fox's front paw is somewhat blurred and less defined than the rest of the body.

Seedream 5.0 Lite

• + Very cute, stylized character design with large expressive eyes.
• + Bright, vibrant colors that enhance the playful and wholesome vibe.
• + Successful inclusion of all requested elements in a clear, central layout.

• − Fur textures look more like 3D-rendered velvet than photorealistic hair.
• − The 'god rays' appear as flat digital gradients rather than atmospheric light.
• − The dew sparkles look like floating white dots rather than reflections on moisture.

GPT Image 2

• + Sophisticated engraving/hatching details on the cloche and banner.
• + Correct Italian orthography for 'Caffè' with the grave accent.
• + Professional composition with a decorative frame that enhances the vintage aesthetic.

• − Slightly busy for a 'minimalist' request, though it fits the vintage brief perfectly.

Seedream 5.0 Lite

• + Clean, minimalist vector style that is very legible.
• + Accurate text for both the brand name and the establishment date.

• − Incorrect accent mark on 'Caffê' (circumflex instead of grave).
• − The cloche and steam elements are very basic and lack the 'vintage' depth requested.

GPT Image 2

• + Excellent typography and nearly perfect text rendering across the entire infographic.
• + Professional, high-quality layout with complex visual elements that remain coherent.
• + Accurate and detailed iconography for the Saturn V and Lunar Module.

• − The style leans more toward detailed illustration than the strictly 'flat-vector' style requested.
• − Slight spelling error on the bottom left (ARMSTRONG is correct, but 'ALDRIN' and 'COLLINS' silhouettes are slightly generic).

Seedream 5.0 Lite

• + Successfully adheres to the 'flat-vector' style with simple, clean shapes.
• + Accurately follows the NASA color palette and the six-step sequence requested.
• + Correct labeling and clear, easy-to-read hierarchy.

• − The 'Translunar' icon is just a gray arc, lacking the requested 'trajectory arc icon' with context.
• − The scale of the Apollo 11 text is slightly overwhelming compared to the small icons.
• − The composition feels a bit sparse with too much empty space.