DALL-E 3 OpenAI vs Z-Image Turbo Alibaba

DALL-E 3

• + High visual quality with intricate wood grain and glass reflections
• + Creative interpretation of the sphere containing a miniature landscape

• − Failed multiple spatial instructions: the book is inside the cube instead of on top
• − The sphere is on top of the book rather than just inside the cube

Z-Image Turbo

• + Perfect adherence to spatial prompts
• + Accurate placement of the red book on top of the glass cube
• + Correct rendering of the small blue sphere inside and a plant behind

• − Lower resolution and more generic aesthetic compared to the alternative
• − The glass cube has a mirrored bottom which wasn't requested

DALL-E 3

• + Excellent atmospheric lighting and puddles with clear reflections
• + Successfully captured 'imperfect framing' with a foreground blur element
• + Strong cinematic quality with high-end photographic composition

• − Anatomical issues in the man's feet and lower legs
• − The bicycle has structural inconsistencies common in AI-generated wheels

Z-Image Turbo

• + Natural skin texture on the man's face and arms
• + Realistic rainfall effect and wet pavement texture
• + The bicycle structure is more coherent than image A

• − Failed to include 'motion blur from passing cars' as cars are static and sharp
• − Composition is very centered and lacks the requested 'cinematic' feel
• − The man is holding the bike rather than actively 'repairing' it

DALL-E 3

• + Excellent high-contrast cinematic lighting with strong orange bokeh.
• + Exceptional detail on the engraved metal and individual beard hairs.
• + Lifelike eyes with intricate iris patterns.

• − The helmet design is somewhat over-engineered and covers more of the head than requested.
• − The skin looks a bit overly polished/airbrushed despite the scars.

Z-Image Turbo

• + Naturalistic and realistic depiction of the dirt and bruising on the face.
• + Braided hair with beads is more clearly defined as requested.
• + The inclusion of the physical torch provides a logical source for the warm lighting.

• − The overall image is slightly softer and lacks the extreme texture detail of the rival model.
• − The bokeh effect is less pronounced and less artistic.

DALL-E 3

• + Provides multiple layout variations in a single image.
• + High-quality, realistic food photography that feels professional.
• + Complex grid structure that fits the minimalist modern aesthetic.

• − Text consists of illegible gibberish symbols.
• − The presentation of four separate pages makes it harder to see one design clearly.

Z-Image Turbo

• + Exceptional text rendering for words like 'APPETIZERS' and 'PIZZA'.
• + Clean, highly organized grid layout that strongly adheres to the prompt.
• + Consistent lighting across all food photos.

• − Spelling error 'MANS' instead of 'MAINS'.
• − The food photography looks slightly more artificial/digital compared to Image A.

DALL-E 3

• + Excellent depiction of the 'exploded' burger concept with distinct floating layers.
• + Captures a high sense of motion and energy with fire and glowing embers.
• + Strong photorealistic textures on the grilled meat and fresh vegetables.

• − Significant spelling errors in the text, including 'MAGIC BURGR' and 'Limiited'.
• − The price tag is rendered in a generic box rather than a starburst as requested.

Z-Image Turbo

• + Perfect text rendering with no spelling errors and a great glowing effect.
• + Accurate interpretation of the 'starburst' for the price tag.
• + High visual quality with realistic food textures and appealing composition.

• − Failed to create an 'exploded' burger, showing a mostly assembled burger instead.
• − Lacks the sense of mid-air suspension for individual components like lettuce and tomato.

DALL-E 3

• + Excellent artistic lighting and atmosphere
• + Realistic chalk texture and decorative flourishes

• − Significant spelling errors throughout the menu
• − Layout is cluttered and difficult to read
• − Failed to render the requested prices accurately

Z-Image Turbo

• + Highly accurate text rendering with almost no spelling errors
• + Clean and legible layout follows the prompt specifically
• + Very convincing handheld chalk handwriting style

• − Slight spelling error on 'Mustroom'
• − Less artistic 'cozy café' atmosphere compared to the other model

DALL-E 3

• + Excellent interior detail with realistic dashboard and lighting.
• + Great atmosphere that captures the feel of Manhattan at night.
• + Creative background detail with 'CAPYBARA' signage on a building.

• − Completely failed to include the human businesswoman in the back seat.
• − The cap is black rather than the requested yellow.

Z-Image Turbo

• + Includes all prompted elements including the human passenger looking at a phone.
• + Accurately colored yellow cap for the driver.
• + Good interpretation of the 'bored' expression on the passenger.

• − Anatomical issues with the capybara's hands which appear primate-like.
• − The lighting is flat and looks like daylight/overcast rather than a vibrant NYC night.

DALL-E 3

• + Exquisite lighting and 3D-depth on the gothic frame elements
• + Strong atmosphere with complex silhouettes of trees and webs
• + High-quality 'cinematic' appearance that feels premium

• − Text is largely unintelligible and includes many typos
• − Failed to render the specific requested address (NYC) correctly

Z-Image Turbo

• + Excellent text legibility and accuracy for almost all requested fields
• + Clear inclusion of the specific scroll banner and holiday details
• + Vibrant colors on the jack-o-lantern are eye-catching

• − The 'Arches' is misspelled as 'Archves'
• − The layout feels more like a digital collage than a cohesive vintage poster
• − Vignette trees and web elements are somewhat generic compared to the other model

DALL-E 3

• + Excellent 3D miniature diorama feel with complex textures
• + Accurate representation of the Japanese flag
• + Clean isometric perspective with a professional finish

• − Failed to place the text 'JAPAN' and 'SUSHI' at the top-center as requested
• − The rice texture looks like large beads rather than grains

Z-Image Turbo

• + Perfect adherence to text placement and content requirements
• + Clean, minimal aesthetic that matches the 'cartoon' request well
• + Accurate 45-degree top-down isometric view

• − Displayed the flag of China instead of the requested Japanese flag icon
• − The diorama base is very simple compared to the '3D miniature' prompt

DALL-E 3

• + Excellent adherence to lighting requests with distinct god rays and dew sparkles
• + Strong artistic consistency and vibrant color palette

• − Fails the hyper-photorealistic requirement by looking more like a digital illustration or 3D animation
• − Strange hybrid 'butterfly-animals' appear in the sky instead of normal butterflies

Z-Image Turbo

• + Successfully achieves a much more photorealistic look as requested
• + Better anatomy and natural integration of the four animals requested
• + Accurate depiction of butterflies and dew drops

• − Lighting is a bit flatter compared to the dramatic god rays requested
• − One of the puppy's paws is oddly positioned/blended into the bunny

DALL-E 3

• + Excellent use of texture and vintage stippling effects.
• + Sophisticated layout with intricate border details.
• + High-quality vector emblem aesthetic.

• − Failed to follow the core text instruction, displaying 'COFFEE HOUSE' instead of 'Caffè Florian'.
• − Layout is more 'complex vintage' than 'minimalist'.

Z-Image Turbo

• + Perfect adherence to text prompt including the specific name 'Caffè Florian'.
• + Closer adherence to the 'minimalist' style requested.
• + Clean, professional typography and layout.

• − The 'Est. 1720' is not on a 'banner' as requested.
• − The steam effect is very simple compared to the artistic curls in the other model.