Black Forest Labs' open-weights image generation model with frontier performance, available for non-commercial local deployment
Settled by community votes across 7 shared challenges, with an AI judge weighing in on each.
FLUX.2 [dev]
#18 of 62 in Text-to-Image
P-Image Edit
#29 of 32 in Image Editing
Where the votes landed
FLUX.2 [dev]
0%
win rate
Ties
0%
P-Image Edit
0%
win rate
Challenge by challenge
The strongest take from each model on every shared challenge, with the AI judge's read.
Man and Car in California
Editing“Make a photo of the man driving the car down the California coastline”
AI Judge Analysis
FLUX.2 [dev]
- + Excellent preservation of the man's clothing, hairstyle, and facial features from the source image.
- + Accurately recreates the interior details and leather seating of the Rolls Royce Phantom Drophead Coupe.
- − The perspective of the car's body in the foreground feels slightly disconnected from the interior view.
- − The lighting on the man's face is quite dark compared to the bright beach background.
P-Image Edit
- + Successfully translates the car from a showroom to a motion shot on a coastline road.
- + Preserves the overall exterior look and angle of the car from the source image very well.
- − The man is barely visible and lacks the specific details (plaid coat, scarf) from the source image.
- − The scale of the man inside the car looks slightly too small.
Verdict: FLUX.2 [dev] performed better as an image editing tool by accurately merging the two subjects, preserving the man's specific attire and hairstyle while placing him inside the car's interior. While P-Image Edit captured the 'driving' action and environment effectively, it failed to maintain the identity and details of the man from the source image, rendering him as a low-detail silhouette.
Pose & Character Mashup
Editing“Use Image 1 as the exact pose reference and Image 2 as the character reference. Recreate the person/character from Image 2 in the exact dynamic pose and body position from Image 1. Keep the exact face, hair, clothing style/details, and expression from Image 2. Match the lighting and environment of Image 1. The final image must show the character from Image 2 performing the precise action/pose from Image 1 with perfect anatomy and natural integration.”
AI Judge Analysis
FLUX.2 [dev]
- + Successfully incorporates the character's clothing, scarf, and sunglasses.
- + Maintains the central yellow background and red ottoman.
- − Extreme anatomical failure with a second head and distorted limbs.
- − Fails to replicate the dynamic pose, opting for a static upright torso superimposed on the scene.
- − Severe artifacting and blending issues between the character and the background elements.
P-Image Edit
- + Captures a much better approximation of the dynamic pose and body position from Image 1.
- + Successfully incorporates the specific character features like the sunglasses and scarf.
- + Integrated the character's face onto the tilted head position with logical orientation.
- − Anatomy is still flawed, particularly with the placement of legs and the extra foot in the air.
- − Confusion in clothing where black sleeves transition into red sleeves.
- − Hair style from Image 2 was not preserved, instead keeping the long hair from Image 1.
Verdict: FLUX.2 [dev] failed the task significantly by creating a nightmarish, multi-headed composite that ignored the requested pose geometry. P-Image Edit, while struggling with clothing consistency and leg anatomy, managed to approximate the dynamic pose and facial orientation of the source images much more effectively. P-Image Edit is preferred for its attempt at structural coherence despite the difficult nature of the pose reference.
Outfit Transfer Challenge
Editing“Use Image 1 as the base person. Dress them in the exact elaborate outfit from Image 2 (including all layers, accessories, jewelry, and shoes). Carefully adapt the clothing to the body shape and pose in Image 1 while maintaining realistic fabric behavior, correct proportions, and perfect lighting/shadow matching. Keep the person’s exact face, hair, and background completely unchanged.”
AI Judge Analysis
FLUX.2 [dev]
- + Successfully captures the texture and pattern of the scarf and pea coat from Image 2.
- + Maintains high resolution and realistic lighting that blends the figure with the beach background.
- − Fails significantly on source preservation by replacing the face and hair of the man in Image 1 with the man from Image 2.
- − Over-complicates the prompt by adding several large gold chains not present in either source image.
P-Image Edit
- + Excellent source preservation, keeping the exact face, unique hair pattern, and skin markings of the person in Image 1.
- + Generally follows the clothing transfer request while adapting the items to the full-body pose.
- − The color of the coat shifted from a dark navy/charcoal to a bright royal blue.
- − The gold chains and yellow-laced shoes are hallucinations not found in Image 2.
Verdict: The task was to keep the person from Image 1 and dress them in the clothes from Image 2. FLUX.2 [dev] failed the primary constraint by replacing the person's face entirely with the subject from the second image. P-Image Edit successfully preserved the original person's identity and background while applying the requested clothing, despite some minor color shifts and added details.
Bald man challenge
Image Editing“Give the person a full, thick head of natural hair with realistic texture, density, and a natural hairline. Preserve facial features and lighting.”
AI Judge Analysis
FLUX.2 [dev]
- + Excellent preservation of the original face, lighting, and background.
- + Realistic hair texture and density that blends well with the existing beard.
- + Maintains the exact geometry of the eyeglasses.
- − The afro-style choice is a bit voluminous, though it fits the prompt's request for 'full and thick'.
P-Image Edit
- + Successfully adds a full head of hair.
- + Preserves the original facial features and environment accurately.
- − The hair texture has 'infinite loop' artifacts where curls repeat unnaturally.
- − The hairline integration at the forehead looks slightly less blending than the other model.
- − The hair rendering looks somewhat more synthetic/plastic compared to the beard.
Verdict: Both models did an excellent job of preserving the source image while adding the requested hair. FLUX.2 [dev] is the winner because the hair texture is more realistic and better integrated with the existing facial hair, whereas P-Image Edit's hair has repetitive digital artifacts in the curls that look less natural.
Over-the-top cartoon caricature
Editing“Create a caricature of me and my job. Make it exaggerated and humorous, incorporating my profession as a tv show anchor and my love for dogs and hockey.”
AI Judge Analysis
FLUX.2 [dev]
- + Excellent caricature style with exaggerated proportions.
- + Effectively integrates all themes including profession (anchor desk), dogs, and hockey simultaneously.
- + Strong character likeness to the source image face.
- − Text on the desk is garbled and unreadable.
- − The hand holding the pen has anatomical issues with the fingers.
P-Image Edit
- + Successfully incorporates the 'tv show' and 'hockey' elements into a cohesive scene.
- + Maintains character likeness while translating the persona into a vibrant cartoon style.
- + Preserves the denim jacket style from the original source image.
- − The 'dogs' requirement is poorly executed, featuring strange, non-organic red shapes on the dog.
- − The overall composition feels a bit cluttered with nonsensical small objects on the desk.
- − Less of a 'caricature' and more of a standard comic illustration.
Verdict: FLUX.2 [dev] produces a much better 'caricature' in the traditional sense, featuring exaggerated heads and a clever integration of humorous elements like hockey-playing dogs. P-Image Edit follows the source material's clothing more closely but fails significantly on the rendering of the dogs, which was a core requirement of the prompt.
Studio Ghibli Anime Style
Editing“Transform this photo into a Studio Ghibli–inspired illustration. Use soft pastel colors, hand-painted textures, gentle lighting, dreamy backgrounds, and a warm, nostalgic mood”
AI Judge Analysis
FLUX.2 [dev]
- + Expertly captures the soft, matte, hand-painted texture typical of Studio Ghibli backgrounds.
- + Excellent application of pastel colors and dreamy, warm lighting that matches the prompt perfectly.
- + Maintains the composition and recognizable poses of the original meme flawlessly.
- − The characters' eyes are a bit simplified, leaning slightly more towards generic anime than a specific Ghibli face style.
P-Image Edit
- + Preserves the facial features and expressions of the source image with high accuracy.
- + Successfully translates the scene into a colorful, dreamy illustrated environment.
- − The art style feels more like a modern digital vector illustration than the requested 'hand-painted' Ghibli style.
- − The lighting is a bit harsher and less 'nostalgic' compared to the first model.
Verdict: Both models successfully interpreted the iconic 'distracted boyfriend' meme into an illustration. FLUX.2 [dev] is the clear winner for its superior adherence to the 'Studio Ghibli' aesthetic, providing the soft textures and hazy, warm atmosphere requested, whereas P-Image Edit produced a cleaner, more standard digital illustration that lacked the specific artistic soul of the prompt.
Golden Hour Stroll
Image Editing“Add dynamic motion to this photo: make hair blow in the wind, add leaves flying, energetic and lively feel.”
AI Judge Analysis
FLUX.2 [dev]
- + Excellent preservation of the subject's face and original features.
- + Realistic, detailed autumn leaves added with natural lighting.
- + Effective wind effect on the hair that looks physically plausible.
- − The leaf placement looks slightly clustered on the left side of the frame.
P-Image Edit
- + Successfully applied the wind effect to the hair and added flying leaves.
- + Maintains high similarity to the original source image structure.
- − The hair edit creates some unnatural, thin 'stringy' artifacts on the right side.
- − Added leaves are very small and look more like simple yellow/orange dots than detailed foliage.
- − Noticeable softening or slight alteration of the facial features compared to the source.
Verdict: Both models followed the instructions well, but FLUX.2 [dev] produced a much higher quality result with more realistic leaf details and a cleaner hair transformation. P-Image Edit struggled with the hair texture and the leaves appeared less detailed, feeling more like graphical overlays than part of the scene.
Explore each model
PrunaAI's sub-1-second multi-image editing model supporting up to 5 reference images with state-of-the-art quality