HiDream AI's image-to-image editing model for instruction-based image modifications and transformations
Settled by community votes across 4 shared challenges, with an AI judge weighing in on each.
HiDream E1
0.7
arena score
#33 of 32 in Image Editing
Skill signature
· Image Editing
Z-Image Turbo
25.3
arena score
#12 of 62 in Text-to-Image
Vote tally
Where the votes landed
HiDream E1
0%
win rate
Ties
0%
Z-Image Turbo
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.”
HiDream E1
Z-Image Turbo
AI Judge Analysis
HiDream E1
- + Successfully added a full head of hair as requested.
- + Maintained the subject's glasses and general jacket style.
- − Completely altered the background, facial skin texture, and lighting.
- − Hair texture looks synthetic and poorly integrated with the original head shape.
- − Significant loss of detail and overall image quality compared to the source.
Z-Image Turbo
- + Perfectly preserved the background, lighting, and facial features.
- + The edited hair matches the original texture and grain of the source image.
- − The hair is quite short/thin and does not fully meet the 'full, thick head' requirement.
- − The subject's glasses were accidentally removed.
Verdict: HiDream E1 followed the prompt's instruction for 'thick hair' but failed significantly as an image editor by completely regenerating the face, clothing, and background, losing all source preservation. Z-Image Turbo acted as a true editor by keeping the original image intact, and although the hair is shorter than requested and the glasses were lost, it is the superior edit due to its high visual fidelity and realism.
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
HiDream E1
Z-Image Turbo
AI Judge Analysis
HiDream E1
- + Successfully incorporated all aspects of the prompt: tv anchor, dog, and hockey.
- + Achieved an exaggerated caricature style as requested.
- + Created a vibrant and humorous composition with hockey players in the background.
- − There is very low source preservation, as the woman's face barely resembles the original.
- − Anatomical issues with the hands holding the dog.
Z-Image Turbo
- + Excellent source preservation, maintaining the woman's actual likeness.
- + Subtly added a small dog in the background.
- − Completely failed to follow the editing instructions regarding the caricature style.
- − Missed the hockey and TV anchor professional elements entirely.
- − Lacks the exaggerated and humorous tone requested.
Verdict: HiDream E1 followed the complex prompt instructions well, successfully integrating the hockey, TV anchor, and dog elements into a caricature style, even though it lost most of the subject's original likeness. Z-Image Turbo largely ignored the edit instructions, providing an image that is almost indistinguishable from the original source except for a tiny dog in the background. HiDream E1 is the winner for actually performing the requested task.
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
HiDream E1
Z-Image Turbo
AI Judge Analysis
HiDream E1
- + Excellent transformation into a stylized illustration style.
- + Strong adherence to the Studio Ghibli aesthetic with clean line work and warm lighting.
- + Maintains the core narrative and composition of the original 'Distracted Boyfriend' meme.
- − Changes the color of the man's shirt from blue plaid to solid pink.
- − Facial expressions are simplified to the point of losing the specific nuances of the original.
Z-Image Turbo
- + Perfectly preserves the identities, clothing details, and composition of the source photo.
- − Fails to follow the 'Studio Ghibli-inspired illustration' prompt entirely.
- − The result is a photo with a slight filter rather than an artistic transformation.
Verdict: HiDream E1 successfully transforms the photograph into a stylized anime illustration that captures the spirit of the prompt, even despite some minor color changes in clothing. Z-Image Turbo almost completely ignores the stylistic request, providing what looks like a slightly desaturated version of the original photo.
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.”
HiDream E1
Z-Image Turbo
AI Judge Analysis
HiDream E1
- + Successfully added very dynamic hair blowing motion
- + Adheres strongly to the request for 'flying leaves' with highly visible elements
- + Maintains vibrant colors and overall composition
- − The added leaves look like stickers and lack realistic integration
- − Significant changes to the subject's face making her look like a different person
- − The dog's face is slightly altered and looks less natural
Z-Image Turbo
- + Excellent source preservation of the woman's face and the dog
- + Integrates the leaves much more realistically with proper scale and lighting
- + Subtle but effective hair movement that feels natural to the scene
- − The 'dynamic motion' effect is a bit too subtle compared to the prompt
- − Hair movement is very slight
Verdict: HiDream E1 followed the instructions for a 'dynamic' feel more literally by adding significant hair movement and large leaves, but it failed to preserve the likeness of the person in the source image. Z-Image Turbo produced a much more realistic edit that perfectly preserved the source characters, though its interpretation of the requested motion was quite conservative.
Next steps
Explore each model
Tongyi-MAI's 6-billion parameter distilled text-to-image model optimized for speed, achieving high-quality generation in 8 steps or fewer with support for bilingual text rendering