GPT Image 1.5 OpenAI vs GPT Image 2 OpenAI

GPT Image 1.5

• + Excellent adherence to lighting instructions with a clear left-to-right gradient.
• + Realistic glass refraction and reflections on the sphere and table.
• + The plant is visible both behind and through the glass cube as requested.

• − The sphere is slightly large for the described 'small sphere' prompt.

GPT Image 2

• + High resolution and crisp texture on the book cover and table wood grain.
• + Clean, modern composition with a very clear glass box.
• + The sphere size better represents the prompt 'small blue sphere'.

• − The lighting is more front-lit than specifically coming from the left window.
• − The refraction of the plant through the glass is less pronounced than in Model A.

GPT Image 1.5

• + Excellent depiction of rain with visible droplets falling and realistic wet-surface reflections.
• + Great bokeh and 50mm feel that matches the cinematic request.
• + Superior skin texture and an 'imperfect framing' that feels like a genuine candid photo.

• − The rear wheel of the bicycle has some structural inconsistencies where the spokes meet the cassette.

GPT Image 2

• + Successfully incorporates motion blur from a passing car in the background.
• + Good rendering of a tool kit and surrounding street elements like the Japanese signage.
• + Accurate elderly facial features and natural hair rendering.

• − The lighting feels flat and less 'cinematic' compared to Image A.
• − The bike's handlebars and frame have significant geometric distortions and floating parts.
• − The rainy atmosphere is very subtle, lacking the texture of light rain requested in the prompt.

GPT Image 1.5

• + Exceptional lighting with realistic warm torchlight reflections on the metal surfaces.
• + Highly detailed cloth and leather textures that meet all prompt requirements.
• + Strong atmosphere with visible bokeh sparks and realistic battle-worn details such as blood and grime.

• − The scars look slightly like surface paint or fresh smeared blood rather than healed tissue.

GPT Image 2

• + More delicate hair braiding and realistic skin texture with subtle freckles.
• + Good use of shallow depth of field for a professional portrait look.
• + Excellent engraving detail on the pauldrons.

• − Lighting is more diffused and lacks the specified 'warm torchlight' intensity seen in the metal reflections.
• − The overall color palette is a bit muted and less cinematic than requested.

GPT Image 1.5

• + Excellent typography rendering with zero spelling errors
• + Clean, high-quality food photography that looks realistic
• + Strict adherence to the bold sans-serif requirement

• − The 'grid' layout for photos is a bit basic, just stacked on the right
• − Less branding elements compared to Model B

GPT Image 2

• + Sophisticated grid-based composition that integrates text and images perfectly
• + Includes additional professional design elements like logos, icons, and a footer
• + Very vibrant use of color and 'vibrant accents' as requested

• − Almost perfect text but contains a minor font artifact on one price ($14.99 looks slightly distorted)
• − Slightly more cluttered than Model A, though still professional

GPT Image 1.5

• + Excellent photorealistic texture on the meat patty and toasted bun
• + Dynamic use of smoke and flying sparks that enhances the fiery theme
• + The text is perfectly integrated and follows the fiery effect prompt

• − The '6.99' price tag has a slight typo in the separator
• − The composition is a bit crowded with the large title overlapping the top bun

GPT Image 2

• + High clarity on individual ingredients like the red onions and dripping sauce
• + Clean and well-organized layout with text on the left and burger on the right
• + The fiery glow effect on the text is very distinct and vibrant

• − The burger feels slightly less 'exploded' and more like a tilted stack
• − The lighting on the lettuce looks a bit more artificial compared to Model A

GPT Image 1.5

• + Excellent text legibility and accuracy of the requested menu items.
• + Consistent chalk texture across all letters.
• + Good use of letter size variation which makes it look authentically handwritten.

• − The composition is a tight crop that misses the 'cozy café' atmosphere requested.
• − The border of the chalkboard is barely visible.

GPT Image 2

• + Stronger adherence to the 'cozy café' setting with background elements and a wooden frame.
• + Includes a small bowl of chalk which adds to the realism of the scene.
• + Text is well-rendered and follows the handwritten request effectively.

• − The title cursive is slightly less elegant than Model A.
• − Minor artifacting in the very small text at the bottom.

GPT Image 1.5

• + Excellent character preservation for the face and specific clothing details like the orange text.
• + Matching the lighting and vibrance of the yellow background perfectly.
• + High visual clarity and resolution in the rendered skin and fabric details.

• − Failed to match the dynamic pose from Image 1, keeping the torso upright instead of bent.
• − Anatomical issues with the feet and how they connect to the legs.

GPT Image 2

• + Successfully replicated the exact tilt and dynamic body position from Image 1.
• + High fidelity in preserving the character's clothing, scarf, and accessories.
• + Excellent integration of the character into the specific lighting and environment of the source image.

• − Minor anatomical artifacts on the lower hand (extra finger/nail detail).
• − The face angle is slightly distorted compared to the reference image.

GPT Image 1.5

• + Excellent cinematic lighting and composition
• + High level of detail in the space background and textures

• − Completely failed the semantic relationship requested in the prompt
• − Standard interpretation of an astronaut riding a horse

GPT Image 2

• + Successfully followed the specific instruction of horse on top
• + High visual clarity and convincing lighting on the subjects
• + Accurate rendering of the NASA logo and space suit textures

• − Anatomical issues where the horse's legs merge into the suit
• − The horse's hooves are holding reins, which is physically nonsensical

GPT Image 1.5

• + Excellent adherence to the outfit details including the watch and sunglasses.
• + High image resolution and sharp texture on the fabric.
• + Successfully replicated the vitiligo pattern on the hands/wrists to match the subject.

• − Substantially altered the person's face and head shape, losing the 'exact face' requirement.
• − Changed the hair style and texture compared to Image 1.

GPT Image 2

• + Perfectly preserved the subject's face, hair, and head shape from Image 1.
• + Accurately transferred the outfit including the plaid scarf and navy coat.
• + Balanced composition that feels very integrated with the original background.

• − Missed the sunglasses from the second image.
• − The lighting on the subject's face is slightly flatter than the original Image 1.

GPT Image 1.5

• + Excellent photorealism with detailed capybara fur and realistic lighting.
• + Accurate adherence to the cap with 'TAXI' text and checkered band.
• + Perfect realization of the woman's bored expression while looking at her phone.

• − The capybara's paws look slightly mutated with too many digits or claws.
• − The perspective through the front windshield feels a bit flat compared to the side views.

GPT Image 2

• + Natural composition with a dynamic angle looking in through the taxi door.
• + Strong bokeh effect on the background city lights enhancing the nighttime atmosphere.
• + High-quality rendering of the capybara's professional posture.

• − The woman's face is slightly blurry and lacks the distinct 'bored' clarity found in the prompt.
• − The 'T' logo on the hat is less iconic than the literal 'TAXI' text requested for a New York theme.

GPT Image 1.5

• + Text rendering is exceptionally clean and perfectly legible throughout the invitation.
• + The thorny border is very prominent and captures the requested gothic texture effectively.
• + Excellent cinematic lighting on the central jack-o-lantern and background graveyard.

• − The composition feels slightly more generic than its counterpart.
• − The banner scroll under the pumpkin is a bit simple in its design.

GPT Image 2

• + Highly detailed and intricate gothic border including skulls and ornate filigree.
• + Includes creative background elements like an actual stone arch bridge, referencing the location 'The Arches'.
• + Superior scroll banner design with more realistic parchment curls.

• − The text at the bottom is noticeably smaller and slightly harder to read compared to the top text.
• − The jack-o-lantern is placed deeper in the shadows, losing some of the 'glowing' prominence requested.

GPT Image 1.5

• + Excellent PBR material rendering on the wood texture and teapot
• + High level of realistic detail on the sushi textures and soy sauce bottle
• + Clean and readable typography with a nicely integrated flag

• − The diorama base is a bit cluttered with large objects like the teapot
• − The isometric perspective feels slightly flattened compared to Model B

GPT Image 2

• + Perfect 45-degree isometric perspective creates a strong 3D diorama feel
• + Creative inclusion of a miniature stone lantern and garden elements
• + Beautifully balanced composition and clean spacing between elements

• − The 'JAPAN' text is slightly less clean with its heavy black outline
• − The chopsticks are rendered with slightly inconsistent thickness

GPT Image 1.5

• + Excellent expressive facial features on all animals
• + Wonderful capture of morning dew and soft particles in the light
• + High level of detail in the fur texture and kitten's paws

• − The kitten has five paws/limbs visible
• − The fox's front right leg is anatomically confusing

GPT Image 2

• + Better sense of motion and 'chasing' as requested in the prompt
• + More realistic flower variety and natural meadow depth
• + Clearer distinction between the individual animals and their actions

• − The fox kit's legs and paws are poorly rendered and distorted
• − The lighting feels slightly more artificial with heavy bloom

GPT Image 1.5

• + Excellent typography with a custom hand-lettered feel
• + Good use of color blocking for a modern-retro look
• + Accurate rendering of the decorative banner

• − Failed the light background instruction by using a solid black background
• − The steam effect is a bit chunky and less elegant than model B

GPT Image 2

• + Perfectly adhered to the light background and subtle texture instructions
• + High-quality vector engraving style with fine detail on the cloche
• + Elegant and balanced composition within a decorative frame

• − The 'EST.' text on the banner is slightly off-center
• − Typography is more standard compared to the creative lettering in model A

GPT Image 1.5

• + Features very clean, bold flat-vector illustrations that match the requested style perfectly.
• + The font rendering is crisp and perfectly legible.
• + The layout uses a comic-style panel approach that is intuitive and engaging.

• − The color of the Earth and the ground plane for Earth Orbit uses a reddish-brown color that feels more like Mars than Earth.
• − Incorrectly spells 'Tranquillity' with an extra 'l' (though this is a valid British variant, it's less standard for NASA contexts).

GPT Image 2

• + Excellent typography and overall poster composition with a clear 'Apollo 11' header.
• + Excellent adherence to the NASA-inspired palette including iconic logos and mission patches.
• + Detailed and accurate icons for each of the six requested steps.

• − The icons for steps 3 and 4 use photorealistic textures that clash with the flat vector style of the other elements.
• − Small text like 'Humanity's first step on the moon' has minor kerning and alignment inconsistencies.