Bing Image Creator vs MidJourney V5: Comparing AI Image Generators
Table of Contents
Introduction to Bing Image Creator and MidJourney V5 Image Generation
The Bing Image Creator and MidJourney V5 are two new AI image generation tools that allow users to create images from text prompts. In this post, we will compare the image generation capabilities of these two tools by testing out a variety of prompts.
The Bing Image Creator is an extension of the DALL-E model developed by Microsoft. It was launched in December 2022. MidJourney V5 is the latest iteration of the MidJourney AI art generator tool, released in November 2022. By testing these systems side-by-side, we can get an understanding of their relative strengths and weaknesses.
Overview of Bing Image Creator and MidJourney V5
The Bing Image Creator provides 10 free daily image generation credits, called 'boosts'. Each prompt allows the user to generate 4 images. MidJourney V5 requires a paid subscription, but offers unlimited generations during an active subscription period. Both tools utilize cutting-edge AI algorithms to generate images from text descriptions. However, they rely on different models and datasets, leading to variances in artistic style, image quality, and prompt interpretation.
Comparing Image Generation Capabilities
In this post, we provide examples of 7 different text prompts provided to both systems. By comparing the output images side-by-side, we assess how accurately each system interprets the prompts, their relative image quality, and their capability to generate creative, high-quality images that match the description. Factors we will compare include adherence to prompt details, inclusion of specified styles/themes, image quality and resolution, creativity and variety of responses, and overall capability to generate complex, artistic images.
Image Prompts and Generation Results
Below we provide 7 text prompt examples, along with the AI-generated images from both the Bing Image Creator and MidJourney V5 for each prompt. This allows us to visually compare how each system interprets and renders the descriptions into final output images.
Beautiful Girl Painting Prompt
For our first prompt, we asked both systems to generate images of a beautiful Sagittarius girl painting in the style of artist Lori Gresely and Studio Ghibli works: The Bing Image Creator output contains elements matching parts of the description, but does not fully capture the specific artistic styles and painting theme requested. The MidJourney results adhere more closely to the prompt with recognizable anime-style portraits of a girl painting.
Healthcare Diving Sunset Prompt
Our next test asks both systems to depict the concept of 'Healthcare diving across the beach at sunset'. This ambiguous prompt aims to test creative interpretation capabilities: While the Bing Image Creator generates valid images of healthcare and sunsets, the interpretations stay quite literal. The MidJourney results take more creative liberty in depicting imaginative healthcare-related concepts diving or floating amidst colorful sunset scenery. MidJourney seems to better embrace imaginative ambiguity here.
Technology Brain Storm Prompt
This prompt asks for an image representing 'Technology brain with storm and lighting' rendered in a 3D digital art style. We want to see interpretation of imaginative anthropomorphism of technology: For this test, the Bing Image Creator produces more abstract, creatively styled depictions of technology/AI concepts surrounded by storms. The MidJourney versions stay closer to literal brains being struck by lighting. We give the advantage here to the Bing Image Creator for its more creative interpretations.
Man in Apocalyptic Cartoon Prompt
Our next prompt asks the systems to generate 'A man in anguish in a realistic, fantastic apocalyptic cartoon style', testing for adherence to multiple descriptive elements: The Bing Image Creator results for this test fail to sufficiently match the descriptive parameters - containing multiple figures rather than a sole man, lacking cohesive apocalyptic themes, and not representing angular facial expressions. The MidJourney results accurately depict a single distraught man against post-apocalyptic backdrops, displaying superior prompt adherence.
Pikachu Workout Prompt
As another test of precise prompt following, we asked both systems to generate images of 'Pikachu doing deadlifts and getting swole at a luxury gym'. While all results contain Pikachu weight lifting, only the MidJourney generations depict the Pokemon character with muscular physiques. MidJourney also provides better gym environment context. This prompt is another win for MidJourney's precise prompt adherence.
Cat Video Game Character Prompt
Testing personification and anthropomorphism, we prompted the AI systems to depict 'A tabby cat styled as an Apex Legends video game character'. For this prompt, the Bing Image Creator produces more normal cat images while the MidJourney results contain imaginative video game character interpretations with matching costume elements. MidJourney shows superior creative styling here befitting the description's intent.
Astronaut on Mars Prompt
As a final test, we asked for 'An astronaut resting on Mars in a beach chair' with vibrant lighting and colors. While both tools generate valid astronaut-on-Mars scenes, the Bing Image Creator better delivers on the specific creative elements - producing an astronaut lounging in a beach chair surrounded by vivid Mars lighting. This round goes to the Bing Image Creator for prompt creativity.
Conclusions on Capabilities
Based on the test prompt results, we can draw some conclusions around the current capabilities of the Bing Image Creator versus MidJourney V5 AI image generation systems.
Summary of Results
Looking at the 7 test prompts and results, MidJourney V5 demonstrated superior capability for precisely matching prompt details, generating images that closely reflect the descriptive parameters provided. However, the Bing Image Creator showed greater tendency to apply creativity, imagination, and interpretation to prompts - sometimes even ignoring specifics to explore concepts in its own style. Out of the 7 rounds comparing output results relative to prompts, we scored MidJourney as 'winning' accuracy and alignment with desired details on 5 prompts. The Bing Image Creator won 2 rounds by demonstrating more abstract, unconstrained creativity freed from prompt limitations.
Recommendations
Based on these exploratory tests, here are some recommendations around how to leverage the strengths of each platform:
- Use MidJourney V5 when you desire output that closely matches detailed descriptive prompts
- Leverage the Bing Image Creator when you want to unleash more open-ended conceptual creativity
- Combine both tools to get a balance of alignment and imagination in your generated images
FAQ
Q: What is the Bing Image Creator?
A: The Bing Image Creator is a new AI image generator from Microsoft built on top of the DALL-E model. It allows users to create images from text prompts.
Q: How does the Bing Image Creator compare to MidJourney?
A: In this testing, MidJourney produced better quality and more accurate images for most prompts. However, Bing Image Creator had some stronger results in areas like realism.
Q: What prompts were tested?
A: Prompts tested included things like painting a girl, healthcare diving sunsets, technology brains storms, apocalyptic cartoons, Pikachu workouts, video game cats, and astronauts on Mars.
Q: Which AI image generator performed better overall?
A: In this limited testing, MidJourney V5 produced better quality and more accurate images for 5 out of the 7 prompts. However, Bing Image Creator shows promise in some areas like realism.
Q: How many free image generations do you get with Bing Image Creator?
A: Bing Image Creator provides 10 free image generation tokens per day. Each prompt uses 1 token to generate 4 images.
Q: What improvements could be made to Bing Image Creator?
A: Bing Image Creator could be improved by enhancing image quality, accuracy, and detail to match prompts more closely like MidJourney does.
Q: Will Bing Image Creator get better over time?
A: Yes, as Microsoft continues to refine the AI model powering Bing Image Creator, we can expect quality and capability to improve significantly.
Q: What does this comparison tell us about AI image generators?
A: This comparison shows AI image generators are rapidly evolving, with new offerings like Bing catching up to more established ones like MidJourney in capability.
Q: Which AI image generator would be better for artwork?
A: Based on this testing, MidJourney V5 would likely produce higher quality, more accurate, and more creative artwork images compared to Bing Image Creator.
Q: Which AI image generator would be better for realism?
A: Bing Image Creator shows early promise in generating more realistic images compared to MidJourney V5 in some test cases.