FLUX.1 Kontext [dev] Black Forest Labs vs FLUX.2 [flex] Black Forest Labs

FLUX.1 Kontext [dev]

• + Excellent adherence to lighting instructions with clear window light from the left.
• + Highly realistic textures on the wooden table and the red book cover.
• + Accurate glass refraction and reflections including the sphere's reflection on the bottom surface.

• − The glass cube lacks a physical back panel, looking more like a frame or open box.

FLUX.2 [flex]

• + Successfully renders the object as a solid glass cube with defined corners.
• + Accurate placement of all elements according to the spatial requirements of the prompt.
• + Clean, modern aesthetic with realistic soft focus on the background plant.

• − The blue sphere has a slightly matte, less realistic texture compared to the other objects.
• − Lighting is a bit flat across the scene despite the window placement.

FLUX.1 Kontext [dev]

• + Excellent depiction of a rainy city atmosphere with wet pavement reflections
• + High clarity and realistic skin texture on the subject

• − The subject is standing with the bike rather than 'repairing' it as requested
• − The composition is too centered and clean, missing the 'imperfect framing' requested in the prompt

FLUX.2 [flex]

• + Strong adherence to the action of 'repairing' the bicycle while kneeling
• + Captures the 'imperfect framing' and 'candid' feel much better with the foreground obstruction and side-profile angle
• + Good implementation of motion blur from passing cars

• − The bicycle's structure is anatomically confusing and lacks a chain
• − Rain effect is less visible compared to Model A

FLUX.1 Kontext [dev]

• + Strong cinematic lighting with high contrast
• + Highly detailed engraving on the golden armor plates
• + Excellent skin texture and realistic eye reflection

• − Missed the request for braided hair with beads
• − The armor looks a bit more like cast bronze than classic plate armor
• − Very little dirt visible on the skin despite the prompt

FLUX.2 [flex]

• + Perfect adherence to specific details like braided hair with beads and visible dirt/scars
• + Excellent texture rendering on leather straps and the knitted cloth underlayer
• + Dynamic composition with visible torch and floating bokeh sparks

• − The scars look slightly superficial/painted on compared to the skin texture
• − The lighting is a bit flat across the face compared to the armor

FLUX.1 Kontext [dev]

• + Excellent high-resolution food photography
• + Strong, bold typography with a modern editorial feel
• + Creative tile-based grid layout

• − Text is largely gibberish and poorly organized as a functional menu
• − Fails to clearly delineate specific sections for Appetizers, Pizza, and Mains

FLUX.2 [flex]

• + Perfect adherence to the structural request with clear Appetizer, Pizza, and Mains sections
• + Includes necessary menu elements like pricing and item descriptions
• + Clean, professional layout that looks like a real-world document

• − Some minor text rendering errors in the smaller descriptions
• − Food photography is slightly less vibrant than Model A

FLUX.1 Kontext [dev]

• + Excellent clarity on the 'Magic Burger' title text
• + Vibrant, high-contrast colors in the background

• − Failed to create an 'exploded' burger with mid-air suspension of parts
• − Contains a typo in the secondary text ('LNHLY' instead of 'ONLY')
• − Starburst icon looks like a flat graphic rather than an integrated fiery element

FLUX.2 [flex]

• + Perfectly adhered to the 'exploded' burger requirement with suspended components
• + Text is rendered with a superior fiery, glowing effect as requested
• + Highly photorealistic textures on the vegetables and meat

• − The '6' in the price has some slight distortion within its shape

FLUX.1 Kontext [dev]

• + Successfully captures a realistic chalk texture on the 'TODAY SPECIALS' text.
• + Includes all requested menu items and prices.

• − Numerous spelling errors including 'Mashroom', 'Risoktso', and 'Octpus'.
• − Major text artifacts and repetitions like 'with with' and dual price tags.
• − The date text is nonsensical and garbled.

FLUX.2 [flex]

• + Excellent prompt adherence with near-perfect spelling across all items.
• + Beautiful chalk texture including realistic background smudges and erasings.
• + Consistently elegant cursive handwriting that matches the requested style.

• − Minor spacing inconsistency on the bottom-most line of text.
• − The word 'SPECIALS' is slightly less detailed in chalk texture compared to the titles above.

FLUX.1 Kontext [dev]

• + Excellent anatomical rendering of the horse and the human face.
• + Clean, cinematic lighting that feels professional.
• + High clarity and sharp resolution.

• − The horse is positioned behind the astronaut rather than literally 'riding' on top of him.
• − The background is quite empty with many simple grey spheres.

FLUX.2 [flex]

• + Perfect adherence to the difficult constraint of the horse riding on top of the astronaut.
• + Vibrant, detailed space background with a nebula and asteroids.
• + Highly creative interpretation that captures the surreal nature of the prompt.

• − The astronaut's hands and the integration with the horse's legs are a bit messy and indistinct.
• − The composition feels slightly more cluttered than Image A.

FLUX.1 Kontext [dev]

• + Excellent photographic lighting and depth of field
• + Highly realistic capybara fur and facial features
• + Captures the bored expression of the passenger perfectly

• − Failed the steering wheel prompt (only one paw on the wheel, the other is on its lap)
• − The car interior feels more like a standard SUV than a specific NYC taxi

FLUX.2 [flex]

• + Adheres better to the 'both front paws on the steering wheel' prompt
• + Better envionmental storytelling with the 'Taxi' light on the roof and clearer Manhattan lights
• + High detail on the taxi driver uniform and cap

• − The scale of the capybara head is slightly awkward compared to the body
• − The passenger is less integrated into the scene lighting

FLUX.1 Kontext [dev]

• + Legible date and time text
• + Includes all requested elements like thorns and bats

• − The scroll text is nonsensical gibberish
• − The location text contains spelling errors
• − The layout feels more like a 2D graphic than a cinematic poster

FLUX.2 [flex]

• + Perfect text rendering for all requested strings
• + Superior atmospheric lighting and depth
• + Highly detailed parchment design with burned edges and subtle skulls

• − The border is slightly less prominent than Model A's

FLUX.1 Kontext [dev]

• + Matches the 'full, thick' instruction with a voluminous hairstyle.
• + Renders hair with high luster and fine strands.

• − Significantly alters the person's face, making them look younger and losing the original's character.
• − Unrelated changes to eyeglasses and beard shape.
• − Lighting on the face is flattened and lost the original's gritty detail.

FLUX.2 [flex]

• + Excellent source preservation, keeping the exact facial features and skin texture of the original man.
• + Preserves the original lighting, clothing, and background perfectly.
• + Adds a realistic, textured buzz-cut style that fits the character's rugged look.

• − The hair is more of a short crew cut than the requested 'full, thick' head of hair.

FLUX.1 Kontext [dev]

• + Excellent clean typography with a designer-like aesthetic.
• + Solid rendering of soft, clay-like textures.
• + Perfect adherence to the solid light blue background request.

• − The 'flag' icon is abstract and does not accurately look like the Japanese flag.
• − The sushi representation is overly simplified and less recognizable than typical sushi.

FLUX.2 [flex]

• + Highly accurate Japanese flag icon.
• + Beautiful isometric composition with multiple varieties of sushi and detailed garnishes.
• + Excellent 3D miniature feel with high-quality PBR-style lighting and shadows.

• − The text is placed lower on the canvas compared to the 'top-center' request.
• − The 'JAPAN' text is simple black rather than having a distinct color treatment.

FLUX.1 Kontext [dev]

• + Successfully converts the person into a bright comic-style caricature.
• + Preserves the original clothing and close-up framing of the source image.
• + Includes the dog and TV elements as requested.

• − Completely misses the 'hockey' requirement.
• − The text 'JOB' and other elements are random and low-effort.
• − Faces and hands within the small screen are poorly defined.

FLUX.2 [flex]

• + Perfectly integrates all elements: news anchor desk, dogs, and hockey gear.
• + Extremely creative interpretation with dogs wearing hockey jerseys.
• + High-quality caricature style that maintains the subject's likeness while adding humor.

• − Slightly changes the pose from a selfie to a seated anchor position, though this fits the prompt better.

FLUX.1 Kontext [dev]

• + Features a beautiful golden-hour lighting with soft bokeh.
• + Clean and stylized visual quality with pleasant facial expressions.

• − Failed the negative prompt by missing several requested animal species (no fox, no bunny).
• − Replaced species with additional kittens of varying colors.
• − Butterflies feel very static and 2D compared to the scene.

FLUX.2 [flex]

• + Excellent prompt adherence including all four specific animals: golden retriever, tabby kitten, bunny, and fox kit.
• + Superior background detail with visible god rays and dew sparkles as requested.
• + Dynamic composition with animals actually in motion.

• − The fox kit has slightly anatomically strange paws.
• − One butterfly wings are disjointed.

FLUX.1 Kontext [dev]

• + Excellent preservation of the source image's composition and poses
• + Clean line work and vibrant colors typical of modern anime
• + High clarity and detail in the clothing textures

• − The style feels more like modern digital anime than the specific 'hand-painted' Ghibli aesthetic requested
• − Misses the soft, dreamy lighting and pastel palette described in the prompt

FLUX.2 [flex]

• + Perfectly captures the 'Studio Ghibli' aesthetic with watercolor-like textures and soft line art
• + Strong adherence to the 'soft pastel colors' and 'dreamy background' instructions
• + Authentic hand-painted paper texture overlay

• − The facial expression of the woman on the right is changed to a smile, losing the narrative 'jealousy' of the source image
• − Some anatomical errors in the man's facial structure

FLUX.1 Kontext [dev]

• + Successfully preserved the subject's identity and clothing from the source.
• + Added wind-swept hair that looks natural and realistic.
• + Included flying leaves as requested.

• − The leaf placement is sparse and looks slightly artificial above the head.
• − The change in dynamics is subtle compared to Model B.

FLUX.2 [flex]

• + Excellent depiction of wind with hair flowing dramatically and upwards.
• + The flying leaves follow a clear path of motion, enhancing the energetic feel.
• + Maintained high fidelity to the original person and dog while significantly increasing the energy level.

• − The dog's tail and the woman's left hand have slight blurring/artifacts from the motion effect.

FLUX.1 Kontext [dev]

• + Perfectly rendered text and accents
• + Clean and bold vector lines
• + Excellent prompt adherence for colors and texture

• − The logo mark looks more like a decorative dome building than a traditional restaurant cloche
• − Missing the requested banner for the 'Est. 1720' text

FLUX.2 [flex]

• + Accurate representation of a restaurant cloche dome on a plate
• + Successfully included the requested banner for the established date
• + Elegant arched typography that feels classically vintage

• − The accent on 'Caffè' is reversed and points the wrong way
• − The steam lines are slightly faint compared to the rest of the logo

FLUX.1 Kontext [dev]

• + Successfully used the specified color palette.
• + Includes iconic representations of astronauts at the end.

• − Significant spelling errors including 'APOLO' and 'DESCNT'.
• − Icons are abstract and messy, failing to clearly represent the Saturn V or the Moon.
• − Layout is cluttered and lacks the logical flow of an infographic.

FLUX.2 [flex]

• + Excellent adherence to the 'flat-vector' style with crisp, clean lines.
• + Legible, perfectly spelled text for all mission stages.
• + Highly accurate iconography representing the Saturn V and Lunar Module.

• − Omitted the final 'Landing' step from the prompt's list.
• − The layout of the trajectory paths is slightly disconnected from the chronological order.