DALL-E 3 OpenAI vs OmniGen v2 VectorSpaceLab

DALL-E 3

• + High resolution with detailed wood grain and glass reflections
• + Captures the lighting mood effectively
• + Creatively interprets the 'blue sphere' with an intricate internal landscape

• − Failed spatial prompt adherence by putting the book inside the cube instead of on top
• − Failed spatial prompt adherence by putting the sphere on top of the book
• − The 'glass cube' is rendered more like a wooden frame with glass panes

OmniGen v2

• + Perfect adherence to all spatial instructions in the prompt
• + Clean, realistic textures for the glass, book, and table surface
• + Accurately renders the plant behind the cube visible through the glass

• − The lighting is a bit flat compared to the requested 'soft window light'
• − The composition is very centered and a bit basic

DALL-E 3

• + Excellent adherence to the 'imperfect framing' and 'candid' aspects of the prompt via foreground bokeh and a side-profile squat.
• + Strong cinematic atmosphere with realistic reflections and depth of field.
• + High level of detail in the bicycle parts and the surrounding Japanese street environment.

• − The man's skin texture and proportions (especially feet) appear slightly more stylized/artificial than requested.
• − There is no visible motion blur on the passing car.

OmniGen v2

• + Natural skin texture on the face and hands lacks the hyper-smooth 'AI' sheen.
• + The red bicycle is a central, clear focal point with accurate proportions.
• + Good representation of light rain and ground reflections.

• − The man is standing with the bike rather than 'repairing' it as requested.
• − The composition is a standard center-weighted shot, ignoring the 'imperfect framing' and 'candid street photo' instruction.
• − The background cars are static and lack the requested motion blur.

DALL-E 3

• + Excellent texture work on the engraved metal and leather straps
• + Strong adherence to the 'battle-worn' aesthetic with realistic scars and lighting
• + Superior cinematic lighting that reflects off the armor convincingly

• − Missed the request for 'braided hair with small beads'
• − The bokeh background is slightly messy with indistinct light blobs

OmniGen v2

• + Followed the specific instruction for braided hair with beads perfectly
• + Clean composition with a very clear shallow depth of field effect
• + Good rendering of the cloth underlayer high on the neck

• − The character does not look 'battle-worn'; the skin is too smooth and the 'dirt' looks like superficial makeup
• − The armor details are softer and less detailed than the other model
• − The 'scars' requested are essentially missing

DALL-E 3

• + Excellent grid composition with dynamic image sizes
• + Includes the specific requested sections like 'MAINS' and 'PIZZA'
• + High visual fidelity for food textures and photography styles

• − Text is largely nonsensical characters and unreadable
• − Layout is presented as a collage of four pages rather than a single cohesive menu sheet

OmniGen v2

• + Stronger adherence to 'bold sans-serif' font request for headers
• + Clear single-sheet menu layout that follows a logical hierarchy
• + Vibrant color accents and professional use of whitespace

• − Frequent spelling errors in main headers (e.g., 'RESTAURATED', 'PIZZZZAN')
• − Food photos have some lighting inconsistencies compared to the background

DALL-E 3

• + Excellent 'exploded' effect with dynamic suspended ingredients.
• + High-quality photorealistic textures on the patty and vegetables.
• + Impressive atmospheric lighting with glowing embers and ground effects.

• − Contains multiple spelling errors in the text ('MAGIC BURGR', 'Limiited').
• − The price tag is not in a starburst as requested.

OmniGen v2

• + Perfect text rendering for 'MAGIC BURGER' and the price.
• + Includes the requested starburst for the price element.
• + Clean, graphic-design style composition.

• − Failed to follow the 'exploded burger' instruction; the burger is fully assembled.
• − Overall visual style is more illustrative/cartoonish than the requested photorealistic detail.

DALL-E 3

• + Excellent chalk texture and realistic lighting
• + Beautiful artistic composition with decorative flourishes
• + Atmospheric cafe setting with warm focal lighting

• − Numerous spelling errors including 'Trufle', 'Occtus', and 'Grililled'
• − Prices are nonsensical and disconnected from items
• − Text structure is cluttered and hard to read

OmniGen v2

• + Text is much more legible and follows the requested menu items more closely
• + Handwriting style is consistent with a chalk marker aesthetic
• + Correctly included the date and approximated the price points for several items

• − The 'chalk' looks more like a digital brush than actual chalk on a board
• − Significant spelling errors such as 'Specals', 'Lemont', and 'Musonhom'
• − Composition is very flat and lacks the cozy cafe atmosphere requested

DALL-E 3

• + Excellent cinematic lighting and atmosphere with a beautiful nebula background.
• + Creative interpretation of 'space' using clouds and stellar objects to ground the scene.
• + Effective use of silhouettes to enhance the surreal feel.

• − Completely failed the negative constraint to have the horse on top of the astronaut.
• − The horse has some anatomical issues with its lower legs and hooves blurring into the clouds.

OmniGen v2

• + Crisp, clean illustration style with clear details on the space suit.
• + Accurate rendering of horse tack like the saddle and bridle.

• − Failed the negative constraint to have the horse on top of the astronaut.
• − The composition is very static and lacks the cinematic quality requested.
• − The floating moons look pasted on and lack integration with the lighting.

DALL-E 3

• + Successfully captures both front paws on the steering wheel while maintaining a professional expression.
• + Exceptional realistic taxi interior details including the dashboard, light fixture, and rearview mirror.
• + Excellent environment consistency with the 'Capybara' sign visible through the window adding a creative touch.

• − Completely missing the human passenger requested in the back seat.

OmniGen v2

• + Includes both the capybara driver and the human passenger looking at a phone.
• + Correctly uses a yellow cap and dark jacket as requested in the prompt.
• + The passenger has the requested bored, nonchalant expression.

• − Failed the steering wheel requirement, showing human hands driving instead of capybara paws.
• − The human passenger is sitting in the front passenger seat rather than the requested back seat.
• − Anatomical failure with a third arm/hand appearing to hold the steering wheel from the passenger's side.

DALL-E 3

• + Exquisite visual quality with a highly detailed, textured 3D gothic aesthetic.
• + Superior composition that feels like a professional, physical vintage poster with layered depths and intricate borders.
• + Strong adherence to the 'cinematic lighting' and 'moody' atmosphere requested.

• − Text rendering is poor, with garbled nonsense letters and incoherent words for the body text.
• − Missing some specific location details like 'The Arches' due to the text issues.

OmniGen v2

• + Text is much more legible and captures large portions of the prompt's specific wording correctly.
• + Follows the layout instructions well, including the scroll banner and the specific location name.

• − The art style is a flat vector illustration rather than the requested 'vintage' or 'cinematic' gothic style.
• − Visual quality is lower with some messy overlapping lines and a generic clipart feel.
• − Text still contains several spelling errors like 'FRIGTS' and 'ARCAS'.

DALL-E 3

• + Excellent 3D miniature diorama feel with high-quality PBR-style textures
• + Creative interpretation of sushi as a stylized isometric block
• + Clean execution of the diorama base and soft lighting

• − Failed to place the text 'JAPAN' and 'SUSHI' at the top-center as requested
• − Rendered the text on the base instead of as floating/top-center elements
• − Missing the 'SUSHI' text word entirely

OmniGen v2

• + Followed text instructions perfectly, placing 'JAPAN' and 'SUSHI' at the top-center
• + Accurate 45-degree isometric projection for both the base and the sushi
• + Very clean, high-clarity image consistent with a 2D/3D cartoon hybrid style

• − The flag icon is incorrect (not the Japanese flag)
• − The sushi rolls have strange structural artifacts (salmon draped over a maki roll with a tail)
• − Textures are more flat and illustrative rather than the requested 'realistic PBR' materials

DALL-E 3

• + Excellent depiction of god rays and morning dew sparkles
• + Fully includes all four requested animals: puppy, kitten, bunny, and fox kit
• + Rich texture in the fur and more natural interaction between creatures

• − The butterfly-animal hybrids in the air are anatomically bizarre and not requested
• − Leans more toward a CGI/3D render than 'hyper-photorealistic'

OmniGen v2

• + Bright, cheerful color palette consistent with a sunrise
• + Clean lines and high clarity for a storybook aesthetic

• − Failed to include the baby bunny requested in the prompt
• − Animals appear like flat 2D illustrations rather than photorealistic subjects
• − Static composition lacks the 'tumbling' and 'chasing' action requested

DALL-E 3

• + Excellent aesthetics with a detailed vintage emblem style
• + Perfectly captures the 'warm brown and cream' and 'subtle texture' requested
• + Strong composition with a balanced hierarchy of design elements

• − Completely failed to use the specific text 'Caffè Florian', substituting 'Coffee House' instead
• − Design is slightly cluttered for a 'minimalist' prompt

OmniGen v2

• + Successfully adhered to the minimalist instruction with clean lines
• + Included the cloche, steam, and banner elements clearly
• + Closer to the requested name, though misspelled

• − Spelling error in the main text ('CAFFFLORIN')
• − Visuals are a bit generic and lack the 'vintage' richness requested
• − Minimalist style feels somewhat under-detailed compared to a professional logo badge

DALL-E 3

• + Excellent adherence to the NASA-inspired color palette and vintage vector aesthetic.
• + High complexity with detailed lunar maps and spacecraft illustrations.
• + Sophisticated layout that feels like a professional museum poster.

• − Includes Space Shuttles which are historically inaccurate for the Apollo 11 mission.
• − The infographic flow is chaotic and does not clearly follow the 6-step prompt sequence.
• − Persistent typos like 'APOLLLO' and 'APPOLO'.

OmniGen v2

• + Strictly follows the requested flat-vector style with consistent iconography.
• + Layout clearly attempts to organize the information into the requested steps.
• + Correct implementation of modern, clean UI/UX design elements.

• − Confuses the mission name, displaying 'APOLO 17' and 'NSA' instead of Apollo 11 and NASA.
• − Text is mostly gibberish or misspelled ('EARTHH ORLDT').
• − Graphic quality is relatively simplistic and lacks the 'NASA-inspired' sophistication.