GPT Image 1 Mini OpenAI vs Qwen Image Edit 2509 Alibaba

GPT Image 1 Mini

• + Successfully merges the exact man and his unique outfit from the source image into the car.
• + Maintains high resolution and realistic lighting that matches the sunset coastline.
• + The car remains very close to the original source model with accurate interior details.

• − The scale of the man relative to the car is slightly too large.

Qwen Image Edit 2509

• + Excellent depiction of the California coastline and cinematic motion blur.
• + The composition and perspective of the car on the road are very natural.

• − Fails to use the specific man from the source image, replacing him with a generic figure.
• − The man is barely visible and the distinct clothing from the source is lost.

GPT Image 1 Mini

• + Successfully integrated character features like the sunglasses, scarf, and black clothing from Image 2.
• + Maintained the yellow background and red ottoman from Image 1.
• + Reasonable face and hair similarity to the character in Image 2.

• − Failed to match the exact dynamic pose, choosing a much simpler kneeling position instead.
• − The anatomy of the feet and legs is slightly awkward.
• − Lost the specific 'sculptural' quality of the pose from the source image.

Qwen Image Edit 2509

• + Perfectly preserved the pose and environment from Image 1.

• − Completely failed to incorporate the character from Image 2.
• − Produced a filtered version of Image 1 without any of the requested edits.
• − Ignored all clothing and facial features of the secondary reference.

GPT Image 1 Mini

• + Successfully replicates the specific plaid pattern and coat style from Image 2.
• + Maintains very high visual quality and realistic lighting/fabric textures.
• + Faithfully reproduces the vitiligo pattern on the hands and face.

• − Changes the person's pose and camera angle significantly rather than editing the source.
• − The face, while resembling the man in Image 1, has been regenerated and is not an exact match.

Qwen Image Edit 2509

• + Correctly preserves the original composition, pose, and background from Image 1.
• + Accurately includes accessories like sunglasses and rings found in Image 2.

• − The face has been significantly altered to look more like the man in Image 2.
• − Poor edge blending where the new head choice meets the original hair line.
• − Lower overall resolution and clarity compared to the other model.

GPT Image 1 Mini

• + The hair has a very natural, matte texture that matches the environment.
• + The transition between the forehead and the added hair is believable.

• − The facial features were slightly altered, making the man look younger and changing his eye shape.
• − The glasses were modified significantly from the source image.

Qwen Image Edit 2509

• + Perfect preservation of the original facial features, glasses, and background.
• + High-resolution hair strands that match the lighting of the scene.

• − The hair style is a bit overly groomed and 'wavy' for the rugged setting.
• − The hair sits slightly high on the head, though still within acceptable limits.

GPT Image 1 Mini

• + Excellent caricature style with great hand-drawn textures.
• + Captures all requested elements: anchor desk, dog, and hockey stick/puck clearly.
• + Maintains strong facial resemblance to the source image while exaggerating features.

• − The composition is a bit flat and static compared to Model B.
• − The news background is very simple.

Qwen Image Edit 2509

• + Very creative composition combining a TV studio with a hockey rink.
• + Strong preservation of the source image's pose and denim outfit.
• + Dynamic 'selfie' perspective that mirrors the original photo's framing.

• − The dog is very small and lacks the 'love for dogs' emphasis requested.
• − Text in the speech bubble and on the monitors is nonsensical gibberish.
• − The face preservation is slightly weaker, losing some of the specific character of the original person's eyes.

GPT Image 1 Mini

• + Successfully captures a hand-painted, soft texture reminiscent of colored pencils or pastels.
• + Maintains the facial expressions and positions of all three characters perfectly.
• + Excellent use of warm, nostalgic lighting that blends the figures into the environment.

• − The color of the foreground dress is still very saturated red rather than Ghibli-style pastel pink/red.
• − The detail in the background buildings is extremely simplified compared to the source.

Qwen Image Edit 2509

• + Clean cel-shaded anime aesthetic that definitely feels like an illustration.
• + Excellent preservation of background details like the red bus and specific buildings.
• + The color palette is lightened and softened to a pastel range as requested.

• − The line art is a bit too sharp and digital to truly feel like a hand-painted Ghibli work.
• − Character faces have been slightly homogenized into a generic anime style, losing some of the specific likeness of the original meme.

GPT Image 1 Mini

• + Excellent integration of flying leaves with shadows on the ground
• + Realistic hair movement and wind-blown texture
• + Maintains strong consistency with the woman's facial features from the original image

• − The dog's tail is slightly altered from the original
• − Large leaves in the foreground are a bit blurry compared to the rest of the scene

Qwen Image Edit 2509

• + Successfully adds highly dynamic, flowing hair effects
• + Leaves are colorful and diverse in shape
• + Preserves the dog and background elements with high fidelity

• − The hair has a slightly 'drawn-on' or plastic appearance with many fine, separate strands
• − Leaves lack shadows and feel somewhat like flat 2D overlays rather than inhabiting the 3D space