Black Forest Labs' open-weights multimodal flow transformer for in-context image generation and editing, available for non-commercial use with character consistency and style transfer capabilities
Settled by community votes across 4 shared challenges, with an AI judge weighing in on each.
FLUX.1 Kontext [dev]
17.1
arena score
#54 of 62 in Text-to-Image
Skill signature
· Text-to-Image
P-Image Edit
9.5
arena score
#29 of 32 in Image Editing
Vote tally
Where the votes landed
FLUX.1 Kontext [dev]
0%
win rate
Ties
0%
P-Image Edit
0%
win rate
Shared challenges
4
Challenge by challenge
The strongest take from each model on every shared challenge, with the AI judge's read.
Bald man challenge
Image EditingEdit instruction
“Give the person a full, thick head of natural hair with realistic texture, density, and a natural hairline. Preserve facial features and lighting.”
FLUX.1 Kontext [dev]
P-Image Edit
AI Judge Analysis
FLUX.1 Kontext [dev]
- + Successfully added a thick head of hair with a realistic hairline.
- + Preserved the lighting and background of the original image well.
- − Significantly altered the facial features, making the person look younger and different from the source.
- − The hair texture looks somewhat airbrushed and less natural than the original beard.
P-Image Edit
- + Excellent source preservation, keeping the face and background identical to the original.
- + Followed the instruction for a thick, full head of hair.
- − The hair texture appears extremely uniform and synthetic, resembling a wig or a preset pattern.
- − The interaction between the hair and the glasses temple is poorly blended.
Verdict: P-Image Edit was far more successful at preserving the identity of the person in the source image, whereas FLUX.1 Kontext [dev] essentially generated a different person with similar features. While P-Image Edit's hair choice looks somewhat like a digital overlay, it creates a much better edit by leaving everything else untouched.
Over-the-top cartoon caricature
EditingEdit instruction
“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.”
Source
FLUX.1 Kontext [dev]
P-Image Edit
AI Judge Analysis
FLUX.1 Kontext [dev]
- + Captures the user's likeness in a comic-like caricature style.
- + Incorporates a dog and a TV showing the anchor profession.
- − Completely misses the 'hockey' requirement.
- − The background elements are flat and less creative compared to Model B.
- − Contains nonsensical text like 'JOB' and garbled letters on the dog tag.
P-Image Edit
- + Successfully incorporates all prompt elements including TV anchor, dogs, and hockey.
- + The composition is highly creative, placing the anchor desk in a hockey arena.
- + Maintains a strong caricatured likeness of the source subject while adding humor.
- − The dog has a strange red heart-shaped structure growing out of its back.
- − The hockey players in the background have slightly distorted faces.
Verdict: P-Image Edit is the clear winner as it successfully integrated all three hobby/job requirements (TV anchor, dogs, and hockey) into a cohesive and humorous scene. FLUX.1 Kontext [dev] completely failed to include the hockey element and provided a much simpler, less imaginative composition.
Studio Ghibli Anime Style
EditingEdit instruction
“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”
Source
FLUX.1 Kontext [dev]
P-Image Edit
AI Judge Analysis
FLUX.1 Kontext [dev]
- + Perfectly preserves the composition and relative positions of all subjects.
- + Highly accurate recreation of the specific clothing patterns and folds.
- + Clean cel-shaded anime aesthetic common in Ghibli films.
- − The lighting feels a bit flat compared to the 'gentle lighting' request.
- − The facial expressions are significantly softened, losing the humor of the original meme.
P-Image Edit
- + Excellent application of soft pastel colors and dreamy, hand-painted textures.
- + Captures the 'warm, nostalgic mood' much better through the floral background and lighting.
- + Preserves the nuanced facial expressions (longing vs. anger) from the source image more effectively.
- − Less precise preservation of the man's plaid shirt pattern compared to Model A.
- − The background is completely replaced with a park rather than a themed version of the original urban street.
Verdict: FLUX.1 Kontext [dev] provides a very accurate technical translation of the photo into an anime style, maintaining the exact urban background and clothing details. However, P-Image Edit better captures the requested 'Ghibli' atmosphere with its superior handle on lighting, textures, and nostalgic mood, despite changing the background context. P-Image Edit is preferred for its artistic adherence to the prompt's stylistic keywords.
Golden Hour Stroll
Image EditingEdit instruction
“Add dynamic motion to this photo: make hair blow in the wind, add leaves flying, energetic and lively feel.”
FLUX.1 Kontext [dev]
P-Image Edit
AI Judge Analysis
FLUX.1 Kontext [dev]
- + Excellent preservation of the original face and clothing details.
- + Subtle and natural-looking hair movement.
- + High overall image clarity consistent with the source.
- − The addition of flying leaves is very sparse and barely noticeable.
- − The dynamic motion effect is relatively conservative.
P-Image Edit
- + Stronger and more energetic application of the flying leaves request.
- + Dramatic wind effect on the hair that conveys more motion.
- + Good preservation of the subject's identity despite the changes.
- − The hair edges suffer from some jitter and loss of detail.
- − The dog's face has changed slightly compared to the source image.
Verdict: P-Image Edit followed the instructions more comprehensively by including a significant amount of flying leaves and more obvious motion in the hair. While FLUX.1 Kontext [dev] maintained higher fidelity to the original textures and the dog's appearance, its interpretation of the edit request was too subtle to truly satisfy the 'dynamic' and 'energetic' requirements.
Next steps
Explore each model
PrunaAI's sub-1-second multi-image editing model supporting up to 5 reference images with state-of-the-art quality