Gemini vs. Imagine with Meta AI Image Generation: A through comparison

User interface and experience

Imagine with Meta and Gemini offer contrasting experiences for generating images through text. Imagine with Meta takes a box-based approach, where users simply type their prompt and receive static image outputs. This focuses solely on the image generation aspect, offering limited interaction or feedback opportunities.

In contrast, Gemini leverages a chat-like interface, creating a more dynamic and interactive experience. Users can refine their prompts, ask clarifying questions, and receive a series of image suggestions based on their ongoing conversation. This conversational approach allows for a more iterative and collaborative image creation process, potentially leading to more tailored and satisfying results.

Watermarking

While both Gemini and Imagine with Meta use AI to generate images from text prompts, their outputs and user experiences differ in important ways. One key distinction lies in their approach to image ownership. Meta embeds a watermark into all generated images, signifying Meta’s claim on the intellectual property. This can be advantageous for users who wish to clearly showcase the AI-generated nature of their work, but also potentially restrictive for creators who desire ownership and commercial use of their prompts’ results.

Gemini, in contrast, currently does not add watermarks. This grants users more freedom to utilize the generated images for personal or commercial purposes without attribution or restrictions. However, this also raises concerns about potential misuse and copyright infringement, as the origin of the images may not be readily apparent.

Prompts to test a generative AI image generator app:

1. Pushing boundaries:

  • Prompt: “A photorealistic portrait of a young woman with cybernetic enhancements, where the technology seamlessly blends with her natural beauty, bathed in the warm glow of a neon city at night.”
  • Evaluation criteria: Look for details in the cybernetic enhancements, realistic human features, and accurate depiction of light and shadow in the neon city environment.
Gemini Imagine with Meta AI
Image of A basic image of a cat, without specific details Image of A basic image of a cat, without specific details
Image of A basic image of a cat, without specific details Image of A basic image of a cat, without specific details
  • Observations: Overall Imagine with Meta AI does a better job visualizing the cybernetic enhancements. Gemini’s cybernetic enhancements do not look as realistic and feels more like some electronic circuits fused into the woemen’s faces.

2. Artistic creativity:

  • Prompt: “A dreamlike landscape painted in the style of Salvador Dali, featuring melting clocks cascading down waterfalls made of honey, with ants crawling across them in the distance.”
  • Evaluation criteria: Assess the level of surrealism, use of melting clocks as a recurring motif, and overall dreamlike atmosphere reminiscent of Dali’s work.
Gemini Imagine with Meta AI
Image of A basic image of a cat, without specific details Image of A basic image of a cat, without specific details
Image of A basic image of a cat, without specific details Image of A basic image of a cat, without specific details
  • Observations: Here the performance of both tools is comparable. Imagine with Meta AI is not able to generate bees correctly. The honey created by Gemini looks more realistic. However, the overall quality and aesthetic of Imagine with Meta AI is paramont.

3. Specific details:

  • Prompt: “A close-up photo of a cappuccino with intricate latte art depicting a dragon breathing fire, served on a rustic wooden table with steam rising from the cup.”
  • Evaluation criteria: Look for precise details in the latte art, realistic coffee texture, steam rising realistically, and accurate depiction of the wooden table surface.
Gemini Imagine with Meta AI
Image of A basic image of a cat, without specific details Image of A basic image of a cat, without specific details
Image of A basic image of a cat, without specific details Image of A basic image of a cat, without specific details
  • Observations: Overall Imagine with Meta AI does a better job, but it does a big mistake in the second image where the dragon is 3D rather than a 2D latte art.

Structured images:

  • Prompt: “an image of a grid with 3 rows and 4 columns. every cell in the grid contains a cartoon image of a creature with a solid white background. The creatures in the grid are as follows: first line: cat, lady bug, pig, goat second line: rooster, dog, lion, butterfly third line: sheep, bee, squirrel, eagle”
  • Evaluation criteria: Assess the generated gird’s structure and the creatures added to the grid, their order, and the background image color as instructed.
Gemini Imagine with Meta AI
Image of A basic image of a cat, without specific details Image of A basic image of a cat, without specific details
Image of A basic image of a cat, without specific details Image of A basic image of a cat, without specific details
Image of A basic image of a cat, without specific details Image of A basic image of a cat, without specific details
Image of A basic image of a cat, without specific details Image of A basic image of a cat, without specific details
  • Observations: Both tools still struggle understanding structured images. Gemini’s first image is a complete miss. However, the second and fourth images are getting the grid’s layout correctly but the creatures are not in order. Imagine with Meta AI can’t get the grid’s structure in any of its generated images but the look and feel of the creatures is closer to the prompt (i.e. cartoon image).

Bonus prompt:

  • Prompt: “A photo of a cat wearing a tiny astronaut helmet, gazing out a window at Earth from the International Space Station.”
  • Evaluation criteria: Assess the cuteness of the cat, accuracy of the astronaut helmet, realistic depiction of the space station interior, and Earth visible through the window.
Gemini Imagine with Meta AI
Image of A basic image of a cat, without specific details Image of A basic image of a cat, without specific details
Image of A basic image of a cat, without specific details Image of A basic image of a cat, without specific details
  • Observations: The performance of both tools is comparable. Both tools are making some minor mistakes. e.g. Gemini is not adding the visor to the second cat’s helmet. Imagine with Meta AI is also sticking the second cat’s ear out of the helmet which does not make sense.