FLUX.2 [flex] Black Forest Labs vs GPT Image 1 OpenAI

FLUX.2 [flex]

• + Excellent adherence to lighting instructions with a clear directional soft light from the left.
• + Highly realistic glass textures and refractions of the interior sphere and background plant.
• + Clean, modern composition with accurate physical proportions.

• − The sphere appears slightly large relative to the term 'small' in the prompt.

GPT Image 1

• + Successfully includes all requested elements in the correct spatial arrangement.
• + The sphere has a nice matte texture that contrasts well with the glass.

• − The glass cube lacks a top pane, appearing more like an open frame which affects the logic of the book sitting on it.
• − The lighting is somewhat flat and doesn't emphasize the 'from the left' instruction as strongly as the competitor.

FLUX.2 [flex]

• + Excellent preservation of the car's model and geometry from the source image.
• + Includes a believable motion blur on the wheels and road.
• + Successful blending of the man into the driving position with a realistic coastal background.

• − The man's clothing was changed from the source image's plaid coat to a white shirt.
• − The man's facial features are slightly simplified compared to the source.

GPT Image 1

• + Successfully retains the man's scarf from the source image.
• + Highly detailed rendering of the man's hair and facial features.
• + Very dynamic and scenic California coastline background.

• − Significant distortion of the car's proportions, making the front grill look squashed and the hood shape inaccurate.
• − The car's door and side profile do not match the source Rolls-Royce Drophead Coupe as well as Model A.

FLUX.2 [flex]

• + Excellent adherence to the 'imperfect framing' prompt with a cinematic obstruction in the foreground
• + Better captured 'motion blur' from passing traffic compared to the other model
• + Dynamic lighting and convincing wet pavement reflections

• − Anatomical issues with the man's pose, specifically his kneeling leg and feet
• − The bicycle structure is physically incoherent, particularly the chain and frame junction

GPT Image 1

• + Exceptional skin texture and realistic facial details
• + More consistent and grounded character interaction with the object
• + A moodier, more naturalistic color palette that fits the 'no stylization' request

• − Fails to include the 'motion blur from passing cars' explicitly requested
• − The bicycle spokes and frame details are slightly messy upon close inspection

FLUX.2 [flex]

• + Exquisitely detailed engravings and leather strap textures
• + Balanced warm lighting with clear torch reflections
• + Very high clarity and sharpness across the subject

• − The facial scars look a bit like fresh red ink rather than healed 'faint scars'
• − The lighting on the face is slightly flat compared to the dramatic armor highlights

GPT Image 1

• + Stronger 'battle-worn' atmosphere with realistic skin texture and grime
• + Exceptional cinematic lighting with deep shadows and soft bokeh
• + More natural integration of hair beads and braids

• − Armor engravings are slightly less distinct than Model A
• − Texture on the underlayer is harder to see due to dark shadows

FLUX.2 [flex]

• + Excellent typography and layout structure resembling a real menu
• + Includes all three requested sections: Appetizers, Pizza, and Mains
• + Sophisticated grid layout for food photography

• − Nonsense filler text for descriptions and some dish names
• − Repeats 'Grilled Chicken Bowl' in two different sections

GPT Image 1

• + High resolution food photography with vibrant colors
• + Very clean and readable sans-serif font
• + Effective use of simple color accents as bullet points

• − Failed to include a dedicated 'Mains' section header as requested
• − Composition is cropped and lacks the full page menu feel of Model A
• − Repetitive placeholder text for every item

FLUX.2 [flex]

• + Perfect text accuracy including the currency symbol and price.
• + Excellent photorealistic textures on the lettuce and bun.
• + Dynamic composition with small particles adding to the motion effect.

• − The starburst effect is a bit generic and clip-art like compared to the rest of the image.

GPT Image 1

• + Very consistent fiery glow across all text elements.
• + Clean, centered 'exploded' layout that is easy to read.
• + Natural integration of the price starburst into the fiery theme.

• − Missing the '6' in the price tag, displaying '.99' instead.
• − Bottom text is slightly cut off at the edge of the frame.
• − The patty texture is slightly less detailed than the competitor's.

FLUX.2 [flex]

• + Excellent text rendering with clean, legible handwriting.
• + Very realistic chalkboard texture with authentic chalk smudges and dust.
• + Successfully interpreted the truncated last item into complete, logical text.

• − The handwriting style feels a bit too uniform, bordering on a digital font look.
• − The lighting on the board is somewhat harsh and uneven.

GPT Image 1

• + Strong 'chalk' feel with granular texture in the strokes.
• + Great variation in letter size and spacing which feels more authentically human-written.
• + Good center-aligned composition that utilizes the vertical space well.

• − Missed the '$' sign on the final item ('Cookies 9').
• − The letter 's' in 'Specials' is slightly malformed compared to other letters.
• − The background context is less defined than the other model.

FLUX.2 [flex]

• + Excellent character preservation of the face, hair, and sunglasses from Image 2.
• + High visual quality with sharp details and realistic skin textures.
• + Maintains the distinct scarf pattern and clothing branding from the character reference.

• − Failed to follow the exact pose from Image 1, opting for a generic crouch instead of the crossed-leg lean.
• − The background lighting is simplified compared to the source studio lighting.

GPT Image 1

• + Successfully replicates the complex crossed-leg pose and body orientation from Image 1.
• + Accurately captures the specific hand and arm positioning of the reference pose.

• − Character resemblance is poor, with a distorted face and mouth.
• − Low visual quality with blurry textures and messy edges around the scarf.
• − The feet and hands have significant anatomical artifacts.

FLUX.2 [flex]

• + Successfully followed the specific instruction for the horse to be on top of the astronaut
• + Vibrant colors and high-quality cinematic background with a detailed galaxy
• + The composition captures the surreal nature of the prompt very well

• − The connection between the horse's hooves and the astronaut's shoulders is a bit awkward and anatomically unclear

GPT Image 1

• + High level of texture on the spacesuit and horse's coat
• + Realistic lighting consistent with a space environment
• + Clean and professional composition

• − Failed the primary logic constraint of the prompt by putting the astronaut on top of the horse
• − The background is less visually interesting compared to Model A

FLUX.2 [flex]

• + Excellent composition that shows more of the taxi and the Manhattan street environment.
• + Very high level of detail in the capybara's fur and the passenger's expression.
• + Captures the professional, focused gaze of the driver perfectly.

• − The driver's paws on the steering wheel look more like primate hands than capybara paws.
• − Slight anatomical confusion where the capybara's ear interacts with the hat.

GPT Image 1

• + Accurate 'TAXI' text on the driver's cap.
• + Very realistic lighting and bokeh in the background city lights.
• + Good adherence to the requested 'bored' expression for the passenger.

• − The capybara's head is disproportionately large for the frame/vehicle.
• − The steering wheel placement looks awkward and physically disconnected from the dashboard.

FLUX.2 [flex]

• + Excellent typography rendering with zero errors.
• + Highly detailed parchment texture with ornate borders including spiders and skulls.
• + Perfect adherence to all requested event details including date, time, and location.

• − The lighting on the pumpkin is slightly disconnected from the atmospheric background.

GPT Image 1

• + Stronger atmospheric and moody color palette.
• + Integration of the pumpkin into the dark environment feels more cinematic.

• − Significant text error where it combined the 'Time' and 'Location' fields incorrectly.
• − The border and parchment texture are less defined compared to the competition.

FLUX.2 [flex]

• + Excellent preservation of original identity and facial features
• + Perfectly maintains lighting and skin texture
• + Natural-looking buzzed/short hair texture that matches the beard

• − The hair is quite short, which may not fully meet the request for a 'full, thick head of hair'

GPT Image 1

• + Provides a very thick and voluminous head of hair
• + Maintains original clothing and background well

• − Significantly alters the person's face and bone structure
• − Hair texture looks slightly matted or less realistic near the hairline
• − The glasses frames were modified and distorted compared to the original

FLUX.2 [flex]

• + Excellent text rendering with clean, bold typography.
• + Precise isometric perspective and framing.
• + Very high clarity with a professional 3D render aesthetic.

• − The flag icon is slightly disproportionate to the text above it.

GPT Image 1

• + Warm, inviting 3D clay-like textures.
• + Good inclusion of extra thematic elements like chopsticks.
• + Clear, legible text.

• − The text color has lower contrast against the background compared to model A.
• − The 'Japan' text is slightly off-center and uses a less impactful font.

FLUX.2 [flex]

• + Excellent integration of multiple elements, showing dogs wearing hockey jerseys in a TV studio.
• + Perfectly captures the subject's outfit from the source image (denim shirt over black top).
• + Clean, modern vector illustration style with high resolution.

• − The facial features move into a generic cartoon territory rather than a literal caricature of the specific individual.

GPT Image 1

• + Authentic caricature style with exaggerated proportions while maintaining the specific facial likeness of the source image.
• + Warm watercolor artistic style that feels very humorous and personal.
• + Effectively includes all elements: news desk, dog, and hockey stick/puck.

• − Text on the background TV is a bit messy compared to the foreground.
• − The hockey integration is slightly less balanced than Model A's thematic jerseys.

FLUX.2 [flex]

• + Excellent adherence to the 'god rays' and 'misty sunrise' lighting description
• + Highly detailed meadow with distinct dew sparkles on the grass
• + Clear representation of all four specific animals with high-quality fur texture

• − The fox kit has black front legs which looks slightly unnatural compared to its other features
• − Composition feels a bit staged with animals lined up

GPT Image 1

• + Dynamic 'tumbling' and 'playful' composition that better fits the energy of the prompt
• + Superior facial expressions that look genuinely joyful
• + Beautiful integration of butterflies within the action of the scene

• − The lighting is very bright, washing out some of the 'dew sparkles' requested
• − The rabbit's scale is a bit small relative to the kitten's paw reaching for it

FLUX.2 [flex]

• + Excellent preservation of the source image's composition and poses.
• + Cleverly translates the 'distracted boyfriend' meme face into an anime style.
• + Captures the clean line work and watercolor-wash aesthetic of Ghibli films like 'Only Yesterday'.

• − The woman on the right appears happy/smiling, which contradicts her original shocked expression.

GPT Image 1

• + Stronger adherence to the requested 'hand-painted textures' with a colored pencil/crayon feel.
• + Maintains the correct emotional dynamic with the girlfriend's angry/scowling expression.
• + Beautiful dreamy color palette.

• − Changes the details of the clothing significantly, particularly the man's shirt pattern.
• − The line work is a bit too soft and lacks the defined character outlines typical of most Ghibli work.

FLUX.2 [flex]

• + Excellent preservation of the subject's face and original clothing details
• + Natural integration of blowing hair that matches the wind direction
• + Adds motion to the dog's tail and posture to enhance the lively feel

GPT Image 1

• + Creates a stronger atmospheric effect with numerous flying leaves
• + Good integration of wind-blown hair that looks voluminous
• + Maintains the overall composition and lighting of the source image well

• − Noticeable change to the subject's facial features compared to the source
• − The dog's tail has a slightly frayed, digital artifact look
• − The leaves appear somewhat repetitive in shape and color

FLUX.2 [flex]

• + Perfectly followed the request for a light background with warm brown and cream tones.
• + Clean vector emblem style with well-executed 'Est. 1720' banner and typography.
• + Excellent use of negative space and minimalist line work consistent with the prompt.

• − The subtle texture requested is almost too subtle to see.
• − The arching of the text is slightly wide compared to the focal point of the logo.

GPT Image 1

• + Strong, visible texture that gives it a nice retro/distressed feel.
• + Accurate typography and banner placement.
• + Clear, bold icon design that scales well.

• − Failed the prompt requirement for a 'light background' by using a solid black background.
• − The 'Est. 1720' text within the banner is slightly off-center and inconsistent in stroke weight.
• − The cloche handle and steam feel slightly more cartoonish than a 'classic' emblem style.

FLUX.2 [flex]

• + Excellent typography with clean, readable text.
• + Accurate and logical flow of mission phases from start to finish.
• + Highly polished vector aesthetic that looks professional and consistent.

• − Misses the final 'Landing' step icon specified in the prompt.
• − The 'Translunar' labels are a bit disconnected from the icons.

GPT Image 1

• + Includes creative extra elements like the crew names and the landing site marker.
• + Captures the NASA-inspired color palette effectively.
• + Incorporates the surface landing icon requested in the prompt.

• − Contains significant spelling errors like 'EARLLUNAR'.
• − Messy layout where labels do not clearly align with their respective icons.
• − The Saturn V rocket graphic overlaps and interferes with the Earth Orbit icon.