HiDream AI's 17B parameter text-to-image model using sparse diffusion transformer with mixture of experts, achieving state-of-the-art image generation quality with strong prompt following
Settled by community votes across 4 shared challenges, with an AI judge weighing in on each.
HiDream I1 Full
#60 of 62 in Text-to-Image
Qwen Image Edit 2509
#32 of 32 in Image Editing
Where the votes landed
HiDream I1 Full
0%
win rate
Ties
0%
Qwen Image Edit 2509
0%
win rate
Challenge by challenge
The strongest take from each model on every shared challenge, with the AI judge's read.
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
HiDream I1 Full
- + Attempts to add hair length
- − Failed the primary instruction by leaving the top of the head bald.
- − Extremely poor source preservation, changing the entire background, facial features, and adding nonsensical artifacts around the beard.
- − Low visual quality with significant hallucinated objects.
Qwen Image Edit 2509
- + Perfectly adhered to the request for a full, thick head of hair.
- + Excellent source preservation, maintaining the original person's face, glasses, clothing, and background flawlessly.
- + Realistic hair texture that blends naturally with the existing beard.
- − The volume of the hair is slightly exaggerated, verging on a wig-like appearance.
Verdict: Qwen Image Edit 2509 successfully completed the edit by providing a thick head of hair while perfectly preserving every other aspect of the original image. In contrast, HiDream I1 Full failed the prompt by keeping the man bald on top and completely changed the person's face and the environment, resulting in a low-quality hallucination.
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
HiDream I1 Full
- + Maintains the facial likeness of the source image much better than Model B
- + Successfully incorporates a dog into a caricature style
- + The caricature style is clean and professionally rendered
- − Failed to include hockey elements clearly in the composition
- − The 'tv show anchor' profession is only vaguely hinted at by a blurred background screen
Qwen Image Edit 2509
- + Excellent integration of all prompt elements including the desk, dog, and hockey players
- + The selfie-style pose from the original image is preserved and adapted to the new setting
- + True caricature style with highly exaggerated features and vibrant colors
- − The facial likeness to the original woman has been significantly altered
- − The text in the speech bubble and on screen is nonsensical/gibberish
Verdict: Qwen Image Edit 2509 is the winner because it successfully integrated every part of the complex prompt—combining a news desk, hockey rink, and dog into a single cohesive scene that felt humorous and creative. HiDream I1 Full produced a higher quality face rendering that looked more like the source subject, but it almost entirely ignored the request to include hockey and failed to place her in a clear professional setting.
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
HiDream I1 Full
- + Successfully captures the character design aesthetic of Studio Ghibli films.
- + Rich, vibrant colors and a charming hand-painted feel in the background.
- + Creative reinterpretation of the scene elements like the floral shirt and Japanese signage.
- − Fails to preserve the original composition and poses of the source image.
- − Lost the dynamic facial expressions (the scowl, the 'pout') that define the meme.
Qwen Image Edit 2509
- + Excellent source preservation, maintaining the exact poses and layout of the original meme.
- + Applied the soft pastel color palette and gentle lighting requested in the prompt.
- + Retained the specific facial expressions of the three main subjects.
- − The illustration style feels more like a standard digital manhwa/webtoon than the specific hand-painted Ghibli style.
- − Background details are a bit washed out and lack the density typical of a Ghibli scene.
Verdict: HiDream I1 Full created a more authentic Studio Ghibli 'look', but it completely failed as an image edit by changing the poses and identities of the characters. Qwen Image Edit 2509 successfully preserved the iconic 'Distracted Boyfriend' composition while applying a soft, illustrative filter that fits the requested mood, making it the better choice for an edit task.
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
HiDream I1 Full
- + Successfully captures a highly energetic and lively feel as requested.
- + Realistic hair movement and clothing dynamics.
- − Failed to preserve the source image; the dog is missing and the woman's identity has changed.
- − Changed the overall lighting and environment excessively.
Qwen Image Edit 2509
- + Excellent source preservation, keeping the woman, the dog, and the background almost identical to the original.
- + Successfully added the requested leaves and hair motion.
- − The wind effect on the hair is slightly unnatural and lacks fine-strand detail.
- − The added leaves look like a flat overlay rather than being integrated into the 3D space.
Verdict: HiDream I1 Full created a more dynamic and 'energetic' image, but it failed as an edit because it completely removed the dog and changed the subject's face. Qwen Image Edit 2509 followed the instructions while maintaining almost perfect source preservation, making it the superior choice for an image editing task.
Explore each model
Alibaba's Qwen image editing model for instruction-based image modifications and transformations