GPT-4o Image Generation vs Midjourney V7: A Comparison 123

Recently, the AI circle has been buzzing with excitement. With the arrival of the all-powerful GPT-4o image generation feature, everyone has been busy getting it to draw. A few days ago, I was scrolling through my WeChat Moments and found that eight out of ten people were showing off Ghibli-style works created by GPT-4o, while the remaining two were still asking for GPT-4o drawing tutorials... The once-silent Midjourney finally couldn't sit still this time. This former "top dog of AI drawing," after more than a year of silence, has made a powerful comeback with its brand-new version Midjourney V7. Let's compare the differences between the two and see which one is stronger.

Feature Comparison

Image Generation Capability:

• MidJourney V7: Overall image quality has greatly improved, with enhanced detail and resolution, significantly reduced artifacts, and an extreme evolution in skin details for character generation. It can better understand and render the complete scene described by prompts, with reasonable spatial relationships between elements, a high degree of integration between background and subject, and the ability to accurately represent specific perspectives, performing professionally in low-angle and close-up shooting styles.

• GPT-4o: Its multimodal image generation feature is powerful, capable of generating images in multiple styles such as the Ghibli style. It only requires two images plus a simple prompt to make the character from the first image appear in the phone mockup of the second image, and can even directly generate photos with alpha channels for convenient post-editing.

Controllability:

• MidJourney V7: Controllability has been improved, but it still retains some unpredictability, which also brings it a unique charm, resulting in more aesthetically pleasing generated images.

• GPT-4o: It offers stronger controllability, generating images more precisely according to the user's prompts. Netizens have developed various functions with it, such as one-click clothing change and image outpainting.

Personalization Settings:

• MidJourney V7: It is the first version to enable model personalization by default. Users need to unlock the personalization settings to use this feature, and the entire process takes about 5 minutes.

• GPT-4o: No similar clearly defined personalization setting feature has been mentioned so far.

Use Case Comparison

• MidJourney V7: Suitable for scenarios with high requirements for image quality and artistic aesthetics, such as artistic creation, advertising design, and fashion magazine cover production.

• GPT-4o: Performs better in scenarios that require quickly generating images in specific styles, conducting creative exploration, and demanding high controllability over images, such as home design, creative poster production, and advertising promotion.

Performance and Efficiency

Image Rendering Speed:

• MidJourney V7: A draft mode has been introduced, costing only half of the standard mode, while increasing image rendering speed by up to 10 times.

• GPT-4o: No specific rendering speed has been mentioned, but based on its functional performance, the speed of image generation is also relatively fast.

Resource Consumption:

• MidJourney V7: The introduction of draft mode has reduced costs, but the specific resource consumption has not been explicitly mentioned.

• GPT-4o: Its resource consumption has not been mentioned so far.

Regarding MidJourney V7, users have expressed recognition of its high-quality generated images and unique artistic style, but some users believe that its insufficient controllability may affect actual work needs. As for GPT-4o's drawing feature, users have praised its strong controllability and diverse functions, but some views also hold that its generated images seem somewhat lacking in artistic depth and emotional expression.