Midjourney v7

Summary for Midjourney v7

Midjourney v7 currently holds an overall Elo rating of 688, placing it in the lower tier of the evaluated models for these specific testing criteria. While it is renowned for its breathtaking aesthetic polish, cinematic lighting, and dense textural details, this evaluation reveals significant limitations in prompt adherence and technical precision.

Key Discoveries:

• Stunning Aesthetics vs. Prompt Adherence: The model frequently scores 8s and 9s for Artistic Merit, but sacrifices accuracy, often ignoring specific prompt details to maintain its preferred cinematic composition.
• Typographical Failures: It struggles profoundly with text generation, consistently producing garbled, misspelled, or structureless typography.
• Stylistic Rigidity: Midjourney v7 applies a highly rendered, 3D/CGI aesthetic to nearly everything, failing to replicate minimalist, flat vector, or traditional 2D animation styles.
• Anatomical Artifacts: Complex interactions often result in merged limbs, extra digits, and incoherent structural geometry.

General Analysis & Useful Insights

Unmatched Textural and Artistic Flourish

Midjourney v7 excels at creating visually arresting imagery. It has a heavy bias toward dramatic lighting, rich color grading, and hyper-detailed surfaces. For example, in the Elaborate facial tattoos prompt, its generated image (View Image) achieved a 9/10 for its exceptional visual presence and cohesive antique metal palette.

The Prompt Adherence Trade-off

The model's biggest flaw is prioritizing its own aesthetic over the user's explicit instructions. In the Cosmic fantasy waterfall prompt, it generated a beautiful standard waterfall (View Image) instead of one made of stars and galaxies. It routinely ignores specific constraints if they clash with its default cinematic style.

Severe Text Generation Weaknesses

With an Elo of 576 in the Text in Images category, Midjourney v7 is not recommended for typographic tasks. Whether it's a Movie Poster (View Image) or a Neon Storefront Sign (View Image), the text is almost always misspelled or completely scrambled.

Anatomical and Structural Hallucinations

When handling specific mechanical or anatomical details, the model frequently falters. In Hand drawing sketch (View Image), it generated an extra digit. In Complex space battle (View Image), ship structures merged into illegible mechanical noise. This limits its utility for tasks requiring precise spatial coherence.

Best Model Analysis by Use Case / Category

Where Midjourney v7 Excels

• Photorealistic People & Portraits (Elo 975): This is the model's strongest category by far. It produces incredibly lifelike skin textures, cinematic lighting, and compelling expressions. The Group Selfie prompt produced an image (View Image) that scored a 9/10 for its rich material differentiation and believable realism.

Use with Caution

• Architecture & Interiors (Elo 577): While the lighting and atmosphere are stunning, the model tends to over-decorate and hallucinate structural elements. In Modern Scandinavian living room (View Image), it ignored the clean minimalist aesthetic, opting for a cluttered, industrial loft look instead.
• Complex Scenes (Elo 620): Great for overall mood and visual spectacle, but individual background elements (like crowds or secondary characters) quickly lose structural coherence.

Where to Avoid Midjourney v7

• Graphic Design (Elo 653): The model cannot do simple, flat vector art. It overcomplicates minimalist prompts, as seen in the Minimalist Coffee Logo (View Image), which was penalized for excessive, tangled linework.
• Ghibli style (Elo 475): The model completely fails to capture traditional 2D animation or watercolor styles, instead rendering prompts like the Spirited Away bathhouse (View Image) in a heavily rendered, CGI-like finish.
• Text in Images (Elo 576): Avoid using this model for any prompt requiring legible, correctly spelled text.