Alibaba's Qwen image editing model for instruction-based image modifications and transformations
Settled by community votes across 7 shared challenges, with an AI judge weighing in on each.
Qwen Image Edit 2511
20.2
arena score
#16 of 32 in Image Editing
Skill signature
· Image Editing
Seedream 4.5
26.1
arena score
#9 of 62 in Text-to-Image
Vote tally
Where the votes landed
Qwen Image Edit 2511
59.8%
win rate
Ties
1.2%
Seedream 4.5
39.0%
win rate
59.8%
1.2% ties
39.0%
Shared challenges
7
Challenge by challenge
The strongest take from each model on every shared challenge, with the AI judge's read.
Man and Car in California
EditingEdit instruction
“Make a photo of the man driving the car down the California coastline”
Source
Qwen Image Edit 2511
Seedream 4.5
70% wins
1% ties
28% wins
AI Judge Analysis
Qwen Image Edit 2511
- + Excellent preservation of the man's facial features and distinctive hairstyle.
- + The interior of the car perfectly matches the source Rolls-Royce aesthetic with cream leather and wood trim.
- + The lighting is vibrant and fits the California coastal theme perfectly.
- − The car is a right-hand drive model, which is atypical for a California-based photo.
- − The man's hands on the steering wheel have some slight anatomical blurring.
Seedream 4.5
- + Successfully incorporates the man's full outfit including the green cargo pants and boots from the source image.
- + Accurately depicts the car driving on the right side of the road with the driver on the left.
- + Strong ocean/cliffside background that clearly resembles the California coast.
- − The man's face and hair are significantly altered from the source image, losing the likeness.
- − Major anatomical error with three legs/feet visible in the footwell area.
- − The car door is open while the vehicle is supposedly in motion, creating a safety and logic error.
Verdict: Qwen Image Edit 2511 is the clear winner as it maintains the subject's identity and facial features with high fidelity, whereas Seedream 4.5 creates a generic face. Although Seedream 4.5 captures more of the subject's clothing, it suffers from severe anatomical glitches and a logical error where the car door is open during driving.
Pose & Character Mashup
EditingEdit instruction
“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.”
Source
Qwen Image Edit 2511
Seedream 4.5
AI Judge Analysis
Qwen Image Edit 2511
- + Retained the original person from the background of Image 2.
- + Followed the dark background aesthetic of Image 2.
- − Failed the primary instruction to replace the subject with the character from Image 2.
- − Overlayed the character from Image 1 on top of the character from Image 2 rather than performing an edit.
- − Poor composite quality with visible cut-outs.
Seedream 4.5
- + Successfully integrated the character from Image 2 into the pose from Image 1.
- + High visual quality and realistic blending of clothing and accessories into a new pose.
- + Matched the yellow environment and lighting of the pose reference accurately.
- − The pose is slightly modified from the original reference to be less contorted.
- − One of the hands is missing a finger.
Verdict: Seedream 4.5 successfully followed the complex instruction to merge the character from Image 2 with the pose from Image 1, effectively adapting the clothing details (scarf, sunglasses, black sweatshirt) to a new dynamic position. Qwen Image Edit 2511 failed the task, essentially creating a poor-quality collage of the two characters standing in front of one another. Seedream 4.5 is the clear winner for its ability to perform high-level character and pose manipulation.
Outfit Transfer Challenge
EditingEdit instruction
“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.”
Source
Qwen Image Edit 2511
Seedream 4.5
AI Judge Analysis
Qwen Image Edit 2511
- + Successfully replicates the entire pose and outfit from the reference image.
- + Generates consistent lighting across the figure.
- − Completely fails to preserve the identity, face, and unique skin features of the person in Image 1.
- − Changes the subjects body type and ethnicity to match Image 2 instead of Image 1.
Seedream 4.5
- + Perfectly preserves the identity, face, hair, and specific vitiligo patterns of the subject from Image 1.
- + Accurately places the outfit from Image 2 onto the subject's body with realistic layering.
- − Leaves an unnatural patch of chest skin visible through the shirt/scarf area that doesn't align with the garment structure.
- − Does not include the sunglasses and shoes requested in the prompt.
Verdict: Qwen Image Edit 2511 failed the core instruction of identity preservation, essentially just swapping the background of the second person rather than editing the first. Seedream 4.5 successfully maintained the specific features of the person from Image 1 while dressing them in the requested coat and scarf, making it a much more successful edit despite minor garment clipping.
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.”
Qwen Image Edit 2511
Seedream 4.5
0% wins
0% ties
100% wins
AI Judge Analysis
Qwen Image Edit 2511
- + Successfully provides a thick, full head of curly hair as requested.
- + Matches the hair lighting and texture to the existing beard.
- + Excellent source preservation of the background and clothing.
- − The hairline on the forehead looks slightly superimposed rather than growing naturally from the scalp.
Seedream 4.5
- + Maintains a very realistic, natural hairline that blends perfectly with the existing temple hair.
- + Preserves the original facial features and environmental lighting perfectly.
- − The hair is quite short and thin compared to the request for 'full, thick' hair.
- − The edit is very subtle, bordering on a buzz cut rather than a full head of hair.
Verdict: Qwen Image Edit 2511 followed the prompt much more effectively by providing a thick, curly head of hair that significantly changes the man's appearance as requested. Seedream 4.5 was very conservative, providing a realistic hairline but failing to deliver the 'thick' volume specified in the instructions.
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
Qwen Image Edit 2511
Seedream 4.5
AI Judge Analysis
Qwen Image Edit 2511
- + Successfully captures a TV news anchor set environment
- + Includes a dog as requested
- + Features an expressive, caricature-style facial exaggeration
- − Completely misses the hockey requirement
- − Nonsensical text and strange candle artifact on the desk
- − The illustration style looks generic compared to the photographic quality of the source
Seedream 4.5
- + Maintains excellent facial resemblance to the source image
- + Successfully incorporates all three elements: TV anchor desk, dog, and hockey equipment
- + Keeps the realistic textures and background elements from the original photo while applying the caricature proportions
- − The hands on the desk are slightly too small for the body scale
Verdict: Seedream 4.5 is the clear winner as it followed all parts of the prompt, including the specific mention of hockey which Qwen Image Edit 2511 missed. Furthermore, Seedream 4.5 maintained better consistency with the source image's visual style and the subject's features while perfectly executing the requested 'big-head' caricature style.
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
Qwen Image Edit 2511
Seedream 4.5
14% wins
0% ties
86% wins
AI judge analysis unavailable for this challenge.
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.”
Qwen Image Edit 2511
Seedream 4.5
14% wins
0% ties
86% wins
AI Judge Analysis
Qwen Image Edit 2511
- + Excellent preservation of the subject's face and original lighting
- + Successfully added a large volume of falling leaves and wind-blown hair
- + The dog's tail was appropriately modified to look fluffier and in motion
Seedream 4.5
- + Natural integration of wind-blown hair that flows to one side
- + Effective use of foreground and background blur on leaves to create depth and speed
- + Adjusted the woman's gait to look more like a jog, enhancing the energetic feel
- − The woman's face has been significantly altered from the source image
- − Background elements like the bridge and trees have changed substantially compared to the source
Verdict: Qwen Image Edit 2511 is the better editing tool as it maintains the identity of the woman and the original composition perfectly while adding the requested elements. Seedream 4.5 creates a more convincing sense of motion through gait changes and motion blur, but it fails as an edit by completely regenerating the subject's face and altering the background environment.
Next steps
Explore each model
ByteDance's latest image generation model unifying text-to-image and image editing in a single architecture, with improved text rendering and 30-40% faster generation than v4.0