Grok Imagine Image xAI vs Wan 2.5 (Preview) Alibaba

Grok Imagine Image

• + Accurate rendering of glass refractive properties and reflections
• + Clean and photorealistic lighting that matches the 'left window' requirement
• + Excellent wood texture on the table surface

• − The blue sphere appears to be floating unnaturally in the center of the cube
• − The cube itself is more rectangular than a true cube

Wan 2.5 (Preview)

• + Excellent texture on the red book cover and pages
• + Pleasant atmospheric lighting with dust particles
• + Good placement of the sphere resting on the bottom surface

• − Perspective issues where the book seems to merge into the top edge of the glass
• − Reflections on the base of the cube are slightly inconsistent with the sphere's position

Grok Imagine Image

• + Excellent photographic quality and motion blur
• + Successfully captures the California coastline aesthetic
• + Highly realistic car design and lighting

• − Completely failed to use the specific man provided in the source image
• − Replaced the subject with an older Caucasian man in sunglasses

Wan 2.5 (Preview)

• + Successfully preserved the identity/appearance of the man from the source image
• + Accurately places the specific car and man into the requested environment
• + Good preservation of the source car's specific details

• − Visual quality is slightly lower than Model A
• − Palm trees in the background look somewhat repetitive/artificial

Grok Imagine Image

• + Excellent depiction of motion blur from passing cars.
• + Highly realistic, candid street photography aesthetic.
• + Perfect adherence to the 'imperfect framing' and '50mm' feel.

• − The man's face is obscured and directed downward, making detail assessment difficult.
• − Rain is less visible than in Model B.

Wan 2.5 (Preview)

• + Clear, high-quality facial details and natural skin texture.
• + Strong atmospheric rain effects and reflections.
• + Solid overall composition and clear subject focus.

• − Missing the requested motion blur from passing cars.
• − The bicycle's rear structure is physically impossible through the wheel.
• − Feels more like a staged high-fidelity stock photo than a 'candid street photo'.

Grok Imagine Image

• + Exquisite engraving detail on the plate armor
• + Beautiful warm lighting and sparks creating atmosphere
• + High-quality rendering of stray hairs and skin texture

• − The beads in the hair are less prominent than requested
• − Armor looks a bit too pristine and clean for a 'battle-worn' description

Wan 2.5 (Preview)

• + Excellent adherence to hair beads requirement with silver accents
• + Great depiction of 'battle-worn' through dirt and frayed cloth textures
• + Strong implementation of leather straps and underlayers

• − Visible AI artifacting/blur on the shoulder plate near the neck
• − The face has a slightly flatter, less lifelike appearance compared to Model A

Grok Imagine Image

• + Excellent structure showing a wide variety of menu items
• + Clean, professional typography with large headers
• + Integrated food photography that feels like a real restaurant menu

• − Has duplicate table entry text like 'Grilled Salmon' and 'Steak Frites'
• − Smaller food images are less detailed and somewhat cluttered
• − Text details are gibberish

Wan 2.5 (Preview)

• + Perfect grid layout for food photography as requested
• + High-quality, vibrant food images with consistent lighting
• + Better adherence to the 'modern minimalist' aesthetic

• − Lacks the volume of menu items found in Image A
• − Headings and sections are a bit sparse for a full menu
• − Text spelling is incorrect for several headers

Grok Imagine Image

• + Excellent typography with a glowing fire effect
• + Very dynamic composition with splashing sauces and embers
• + Strong photorealistic textures on the lettuce and burger patty

• − The starburst for the price is a bit flat compared to the rest of the image

Wan 2.5 (Preview)

• + Creative melting cheese effect integrated into the main title typography
• + Very clean center-focused composition
• + Excellent fiery effect on the price starburst

• − The tomato slices look slightly more CGI than photorealistic
• − The 'Limited Time Only' text is slightly less readable than Model A

Grok Imagine Image

• + Exceptional text rendering with no spelling errors.
• + Realistic chalk texture with dusty smudges and subtle grain.
• + Perfect adherence to the requested items and pricing.

• − The 'elegant cursive' for the title is more of a script than formal cursive.

Wan 2.5 (Preview)

• + Strong chalk-like aesthetic with natural-looking underlines.
• + Excellent layout that feels authentic to a café atmosphere.

• − Missed the word 'Herbs' in the second menu item.
• − Repeating prices and awkward line breaks in the cookie section.
• − The handwriting looks slightly more digital/clean than Model A.

Grok Imagine Image

• + Successfully followed the specific instruction for the horse to be on top of the astronaut
• + Rich, vibrant colors in the nebula background with a high sense of surrealism
• + Good anatomical detail on the horse

• − The connection between the horse and astronaut is a bit floaty rather than a 'riding' pose
• − Some minor artifacting where the front hoof meets the astronaut's hand

Wan 2.5 (Preview)

• + Excellent visual clarity and sharp details on the astronaut suit and horse hair
• + Great composition with the earth and galaxy in the background
• + Cinematic lighting and sense of motion

• − Failed the negative constraint/specific instruction by placing the astronaut on top of the horse
• − The ground texture at the bottom feels slightly inconsistent with the space setting

Grok Imagine Image

• + Successfully preserved the exact person, face, and unique skin features from Image 1.
• + High resolution and intricate detail in the clothing applied.
• + Perfectly preserved the background and wooden structure.

• − Completely failed to use the specified outfit from Image 2, instead generating a generic elaborate costume.
• − Ignored the specific modern coat, scarf, and sunglasses in Image 2.

Wan 2.5 (Preview)

• + Successfully used the exact outfit, scarf, sunglasses, and watch from Image 2.
• + Applied the clothing to the pose of the base image very effectively.
• + Maintained the background lighting and environment from Image 1.

• − Completely failed to preserve the person from Image 1, replacing the face and skin with the man from Image 2.
• − Failed to keep the person's exact face and hair as requested in the instructions.

Grok Imagine Image

• + The lighting and textures on the capybara are highly realistic.
• + The woman is sitting in the passenger seat as specified in the prompt.
• + The interior details, like the dashboard and rear-view mirror, are very convincing.

• − The woman appears to be sitting in the front passenger seat rather than the back seat.
• − The taxi light on top of the car is visible through the roof, which is a structural impossibility.

Wan 2.5 (Preview)

• + Successfully places the passenger in the back seat as requested.
• + The capybara's hat is more detailed and resembles a formal driver cap.
• + The rainy atmosphere and city lights through the window add great aesthetic value.

• − The text on the taxi sign is nonsensical.
• − The steering wheel appears to be melting into the capybara's paws.

Grok Imagine Image

• + Excellent gothic typography that feels integrated with the theme
• + Strict adherence to all prompt elements including the thorn and web border
• + Strong atmospheric composition with cinematic lighting around the jack-o-lantern

• − The parchment edges look slightly clipped against the black background

Wan 2.5 (Preview)

• + Highly detailed jack-o-lantern with a dynamic fire effect
• + Vibrant colors and high-quality textures on the twisted tree and scroll
• + Perfect spelling and legible event details

• − The 'Halloween Party Invitation' text lacks the requested 'gothic' style, looking more like standard 3D serif font
• − The blue sky background feels a bit too comic-like compared to the requested 'dark parchment' vintage look

Grok Imagine Image

• + Excellent preservation of original facial features and bone structure
• + Hair texture and color match the existing beard perfectly
• + Highly realistic hairline and integration with the ears

• − The forehead height remains slightly high, giving a receding look despite the prompt for full hair

Wan 2.5 (Preview)

• + Very thick and voluminous hair that clearly satisfies the 'full' requirement
• + Matches the facial lighting well

• − Slightly altered the person's eye shape, making him look like a younger/different person
• − The hair volume is a bit stylized and looks less 'natural' compared to the reference beard texture
• − Minor distortion on the ear-stem of the glasses

Grok Imagine Image

• + Excellent adherence to the 45° isometric perspective including the base.
• + Very clean typography and flag placement as requested.
• + Realistic PBR-style textures and lighting on the wood and ingredients.

• − The white nigiri has a slightly floating appearance relative to the rice.

Wan 2.5 (Preview)

• + Beautiful soft-render 3D aesthetic and high-quality textures.
• + Correct text spelling and inclusion of the flag element.
• + Nice focus and depth of field effect.

• − Failed the isometric 45° perspective requirement, opting for a standard portrait angle.
• − The flag icon is positioned to the side rather than the requested top-center location.

Grok Imagine Image

• + Retains a higher level of facial likeness from the source image compared to the other model.
• + Successfully integrates all elements: TV anchor desk, hockey rink background, and a dog in hockey gear.
• + Clean, high-quality digital art style that fits the 'humorous caricature' request perfectly.

• − The hands are slightly malformed and undersized for the character's body.
• − The pucks in the background look a bit generic and flat.

Wan 2.5 (Preview)

• + Incorporates the blue denim shirt from the source image as a nod to the original outfit.
• + Clearly includes hockey items (stick, jersey) and multiple dogs in the composition.
• + Strong caricature 'big head' style with exaggerated features.

• − The facial likeness is significantly lower, making the subject look like a generic cartoon character.
• − The background TV shows what appears to be a hockey player on a grass field (likely field hockey/football hybrid).
• − Visible artifacts on the hockey stick and the microphone.

Grok Imagine Image

• + Excellent depiction of golden hour lighting and sun rays
• + Cute, stylized character designs that fit a wholesome vibe

• − Artificial, 'AI-illustrated' look with painted fur textures
• − Failed to include butterflies mentioned in the prompt
• − Poorly defined backgrounds and anatomical merging of the animals

Wan 2.5 (Preview)

• + Much higher realism in fur textures and animal anatomy
• + Successfully included all elements, including butterflies and dew drops
• + Superior composition with a more natural depth of field

• − The fox kit has slightly uncanny, human-like eyes
• − Some floating artifacts or seeds in the air look a bit messy

Grok Imagine Image

• + Excellent hand-painted, watercolor-like texture that feels authentic to Ghibli background art.
• + Accurately replicates the character positions and expressions while translating them into an anime style.
• + The background buildings and sky are beautifully reimagined in a dreamy, nostalgic style.

• − The man's facial features (beard and eyes) feel a bit more like generic illustration than the specific Ghibli facial archetypes.

Wan 2.5 (Preview)

• + Strong character designs that very closely mimic the Ghibli 'hero/heroine' face shapes and eyes.
• + Great use of 'dreamy' elements like floating leaves and glowing dust motes.
• + Preserves the composition of the original photo perfectly.

• − The textures are a bit too smooth and digital, lacking the requested hand-painted or watercolor feel.
• − The colors are somewhat washed out compared to the vibrant but soft palette expected of Ghibli.

Grok Imagine Image

• + Excellent adherence to the 'energetic and lively' instruction with many flying leaves.
• + Effective wind effect on the hair that looks natural and broad.
• + Great preservation of the original subjects and background elements.

• − The orange autumn leaves clash slightly with the green trees in the background.
• − Some leaves overlapping the subject appear a bit sharp and flat.

Wan 2.5 (Preview)

• + Natural-looking wind effect on the woman's hair.
• + High image quality and preservation of the original scene details.
• + Subtle and clean editing of the leaves.

• − The green leaves look like stickers and lack natural depth or motion blur.
• − Fewer leaves than Model A makes the scene feel less 'energetic'.

Grok Imagine Image

• + Excellent typography rendering for both the main name and established date.
• + Clean vector emblem style with modern minimalist appeal.
• + High contrast and sharp details on a textured light background.

• − Redundant text as it includes 'Est. 1720' twice.
• − Integration of a spoon/handle into the cloche dome is slightly messy and unconventional.

Wan 2.5 (Preview)

• + Cohesive use of framing and vintage paper textures for a 'vintage' feel.
• + Elegant banner design that flows well with the cloche graphic.
• + Classic typography that matches the historical aesthetic perfectly.

• − The 'Est.' text is slightly less crisp compared to the main title.
• − The steam vector is a bit thinner and less prominent than in Image A.

Grok Imagine Image

• + Successfully included all 6 numbered steps requested in the prompt.
• + Followed the color palette requirements more closely with prominent navy and muted red.
• + Text rendering is remarkably clean and legible for an AI generation.

• − Step 3 text has gibberish/artifact words ('3rajcoory', 'Transluiory').
• − The Saturn V icon is a bit simplified/stylized compared to a realistic vector.

Wan 2.5 (Preview)

• + High-quality vector illustrations for the Saturn V and Lunar Module.
• + Excellent use of space and logical flow for the trajectory lines.
• + Clean, professional aesthetic suitable for a real poster.

• − Failed to include 6 distinct numbered steps as requested in the instructions.
• − Steps 5 and 6 (Descent and Landing) are just text labels without dedicated icons.
• − Text alignment for the steps is somewhat disorganized compared to Model A.