GPT Image 1.5 OpenAI vs Stable Diffusion 3.5 Large Stability AI

GPT Image 1.5

• + Perfect adherence to the spatial arrangement requested.
• + Highly realistic glass reflections and refractions of the background plant.
• + Excellent material textures, especially on the book's canvas cover and the wooden table.

• − The blue sphere is relatively large compared to the prompt's 'small blue sphere'.

Stable Diffusion 3.5 Large

• + Good lighting effects and sharp focus on the glass cube.
• + Accurate 'small' scale for the blue sphere.

• − Incorrect object placement; the book is inside the cube rather than sitting on top of it.
• − Coherence issues where the glass cube seems to clip through the red book.
• − The plant is mostly to the side/front rather than behind the cube as requested.

GPT Image 1.5

• + Excellent skin and fabric textures that look realistic.
• + Strong adherence to 'imperfect framing' with a tight, candid composition.
• + Detailed mechanical parts on the bicycle and repair tools.

• − The car in the background lacks the requested motion blur.
• − The raindrops appear as static white dots rather than falling streaks.

Stable Diffusion 3.5 Large

• + Successfully captured motion blur on the background vehicles.
• + Good representation of falling rain and wet pavement reflections.
• + Accurate adherence to 'shallow depth of field' with a soft background.

• − Anatomical issues with the man's hands and arms which appear distorted.
• − The bicycle's structure is physically impossible with floating and merging parts.
• − The man appears slightly 'pasted' into the scene with mismatched lighting.

GPT Image 1.5

• + Excellent adherence to all prompt details including beads in hair and leather straps.
• + Superior lighting effects with warm torchlight and realistic bokeh sparks.
• + Highly detailed facial texture with convincing scars and dirt.

• − The composition is a bit tight on the forehead.

Stable Diffusion 3.5 Large

• + Beautifully detailed ornate engraving on the plate armor.
• + Strong character expression and clear facial features.
• + Good interpretation of the braided hair requirement.

• − Missed the 'small beads' in the hair mentioned in the prompt.
• − The lighting feels more like daylight than the requested warm torchlight.
• − Lacks the specific bokeh sparks requested.

GPT Image 1.5

• + Excellent text rendering with all requested strings accurate.
• + Perfect adherence to the 'exploded' layout requested.
• + High photorealistic texture detail on the patty and bun.

• − The lighting is very intense, bordering on busy.

Stable Diffusion 3.5 Large

• + Great visual quality on the food textures.
• + Dynamic flame effects around the base of the burger.

• − Failed to include any of the requested text.
• − The burger is not 'exploded' as requested, but mostly assembled.
• − Missing the required starburst element.

GPT Image 1.5

• + Excellent text rendering with no spelling errors
• + Highly realistic chalk texture and natural handwriting style
• + Perfect adherence to the specific date and menu items provided

Stable Diffusion 3.5 Large

• + Provides a wider architectural context of a cozy café environment
• + Good composition with greenery and lighting
• + Captures the general aesthetic requested

• − Numerous spelling errors including 'TODAAY' and '2024' instead of '2026'
• − Text is cluttered and doesn't follow the specific menu item list correctly
• − Uses a mix of styles that look more like digital fonts than natural handwriting

GPT Image 1.5

• + Excellent high-contrast textures on the horse and suit
• + Solid compositional layout with multiple space elements
• + Sharp rendering of fine details like the lunar lander and asteroid dust

• − Failed the negative constraint to put the horse on top
• − The horse's front-right leg has an anatomical glitch at the joint

Stable Diffusion 3.5 Large

• + Successfully followed the difficult spatial constraint of putting the horse on top of the astronaut
• + Ethereal, dreamlike color palette and lighting
• + Good sense of motion and scale against the planet's curve

• − The horse's head and neck anatomy are slightly distorted
• − Lower overall sharpness compared to the other model

GPT Image 1.5

• + Excellent adherence to the full prompt including the passenger.
• + Photorealistic lighting and textures that feel grounded in reality.
• + Accurate capybara anatomy and driver accessories.

• − The passenger's face is slightly blurry.
• − The paws on the steering wheel look slightly human-like in their grip.

Stable Diffusion 3.5 Large

• + High saturation and clean, modern colors.
• + Detailed rendering of the capybara's fur and clothing.

• − Completely missed the secondary character in the back seat.
• − The capybara's head shape and ears look more like a rat or degu than a capybara.
• − Logical inconsistencies with the steering wheel placement and the animal's limbs.

GPT Image 1.5

• + Excellent text rendering with no spelling errors across all required fields.
• + Highly cohesive gothic aesthetic with a beautiful thorn and web border.
• + Superior cinematic lighting and texture on the central jack-o-lantern.

• − The parchment texture is slightly less distinct than the torn paper edge look.

Stable Diffusion 3.5 Large

• + Creative use of a torn parchment edge as a frame.
• + Dynamic composition with a large, glowing moon.

• − Failed to include the specific event details (Date, Time, Location) at the bottom.
• − Text rendering on the scroll is messy with artifacts and incomplete words.
• − Composition is cluttered, making the 'Party' text feel squashed.

GPT Image 1.5

• + Excellent typography rendering with clean, centered alignment.
• + High-quality PBR textures for the wood, ceramic, and food items.
• + Perfect adherence to the 45-degree isometric diorama perspective.

• − The lighting is a bit more contrasty than the 'gentle' request imply.

Stable Diffusion 3.5 Large

• + Features a wider variety of sushi pieces.
• + Good textures on the salmon and rice grains.

• − Failed to place the text at the top-center, putting it on a small sign instead.
• − The chopsticks are poorly rendered and overlapping incorrectly.
• − The perspective is not a true isometric 45-degree angle.