Have you ever wondered which artificial intelligence generates the best images? I tested three incredible models - Leonardo AI, Luma Photon and Google Imagen 3 - to find out which one really stands out.
Comparing AI-generated images can be quite subjective, but I've developed an innovative method that makes this analysis easier: the Characteristic Breakdown Method (CBM).
CBM is a systematic approach to evaluating the quality of images generated by different AI models.
It works by dividing the original prompt into several distinct characteristics and analyzing the performance of each model based on these characteristics. This makes the evaluation more objective and structured. Check out the full verdict throughout the article!
Prompt used to compare three different image-generating models:
"An Unsplash-style portrait of a creative professional in a minimalist office. The background is gently blurred with a bokeh effect, while the subject wears a neutral outfit accented by a delicate purple accessory, such as a scarf or wristwatch, adding a unique yet understated pop of color."
Elements of the prompt:
- Unsplash-style portrait
- Creative professional
- Minimalist office
- Softly out of focus with bokeh effect
- Neutral attire
- Delicate purple
- Scarf or wristwatch
- Subtle touch of color
- Clean and professional aesthetics
- Soft and natural
- Modern, sophisticated
- Minimalist and clean
Using the CBM Decomposition Agent, I generated a complete table, highlighting the following characteristics:
- Style
- Occupation
- Configuration
- Fund
- Color of clothing
- Accessory color
- Type of accessory
- Color Strategy
- Composition
- Lighting
- Humor
- Visual texture
Dividing the prompt into 12 characteristics automatically with the CBM makes it much easier to assess the quality of the models.
The results of the comparison between the three AI models - Leonardo AI, Luma Photon and Google Imagen 3 - showed significant differences in the quality of the images generated. See below for details of these differences!
Leonardo AI
Leonardo AI stood out for its consistency and precision. It managed to maintain high fidelity to the original prompt in almost every feature, from the style and configuration to the color of the accessory and the lighting.
This resulted in a very high final score, reflecting its ability to generate images that exactly match the user's expectations.
- Technical score: 97.30%
- Quality Score: 100.00%
- Final score: 98.65%
Luma Photon
Luma Photon performed similarly to Google Imagen 3. It performed well in several characteristics, but also struggled in areas such as the color of the accessory and the continuity of soft, natural lighting. Its final score was slightly lower than that of Google Imagen 3.
- Technical score: 97.30%
- Quality Score: 90.00%
- Final score: 93.65%
Google Image 3
Although Google Imagen 3 also generated high-quality images, it struggled with some specific features, such as the bokeh background and color strategy.
This resulted in a lower final score compared to the Leonardo AI, but it still proved to be a competent model.
- Technical score: 82.43%
- Quality Score: 80.00%
- Final score: 81.22%
Conclusion
All in all, Leonardo AI excelled in most features, offering a more consistent imaging experience that was faithful to the original prompt.
So, do you agree with the results? Are you satisfied with the performance of each AI? Would you like to explore new models?
We invite you to try the Tess AI platform and discover the power of Leonardo AI, along with the other incredible AIs available on the platform.
Explore a variety of imaging tools, all in one place, without the need for multiple subscriptions!
