Unlocking the Potential of Leonardo AI: A Game Changer for Image Guidance
Written on
Chapter 1: The Edge of AI Image Generation
In the rapidly evolving realm of AI, Leonardo AI’s Image Guidance is proving to be a significant advancement. This week, it's safe to say that it stands out as a real "game changer."
Take a moment to examine the image provided and reflect on which of the three might be an AI-generated image and which is definitely a photograph. After sharing this with friends, it was surprising to see that everyone struggled with the answers. However, the true marvel of this exercise lies not in the realism of the images, but in how closely the AI-generated versions resemble the actual photograph.
In this article, I will delve into the IMAGE GUIDANCE feature offered by Leonardo AI, which allows for the creation of highly accurate images and prompt generation inspired by traditional photography. Suppose you have photographs from a conventional shoot where the model's expression or pose wasn't ideal. With Leonardo AI, you can bring in that image, apply Image Guidance, and refine it to your satisfaction using the same model. This process goes beyond simple facial adjustments and certainly surpasses the seed function available in Midjourney.
Moreover, this feature enables the creation of compositions that offer unique opportunities and aesthetics, far beyond what conventional prompts can achieve. The more lifelike the original photograph, the more realistic the outcome from Leonardo AI. In a prior article, I integrated both Midjourney and Leonardo AI for some stunning outcomes, yet the final results still felt somewhat artificial.
Chapter 2: Exploring the Capabilities of Image Guidance
As you explore the differences between the initial photograph I used for Image Guidance and the final results, consider whether you managed to answer my earlier questions correctly.
In this scenario, the overall lighting improved significantly, the colors were refined to appear more natural, and the facial features were softened while still maintaining realism. It's fascinating to note the similarities between the central image and the one adjacent to it—the patterns in the rattan chair and the creases in the cushion, as well as the wear in the jeans, are nearly identical. The side images reveal that the center image appears less convincing, particularly in terms of lighting. Thankfully, Leonardo handled all the color adjustments seamlessly!
How the Image Guidance Feature Works
The interface of Leonardo AI includes a text prompt section at the top, but take note of the green IMAGE GUIDANCE button—it’s activated. This feature allows you to upload up to four images for guidance. For instance, if you aim to "reshoot" a specific model with remarkable accuracy, you can upload four guidance images and then set your new prompt. It’s crucial to maintain a level of coherence between the imagery and the new text prompt. For instance, if you have images of a male chef in a kitchen, it would be quite challenging to transform them into a female firefighter on the moon. However, making general adjustments to expressions, poses, and locations is fairly straightforward for Leonardo AI when Image Guidance is enabled.
Allow me to illustrate the capability of this system with another example. Here’s the prompt I used:
Prompt: ‘young hipster man in a white t-shirt holding a brown pitbull puppy up to his face, smiling at the camera, in a park, tattoos on arms, side view, energetic, shot with Canon R5, 70–200mm, cinematic studio lighting, 8k UHD, photographed by Bert Stern.’
The image produced from this prompt was decent, but nothing particularly impressive compared to what Midjourney could achieve.
Now, observe the level of detail and realism achieved using the Image Guidance feature.
These images not only resemble the original closely but also exhibit far greater realism than simply relying on a text prompt. Notice the similarities between the original image and the one on the right, as well as the subtle variations in the man's facial expression on the left. This is attributed to the “Strength” parameter of the Image Guidance, which can be adjusted from 0 to 1.0, with notable differences in output depending on the value set.
Consider the following images produced from another prompt:
Prompt: ‘photorealistic commercial fashion photo of a 21-year-old woman in a vibrant bohemian bikini top and patterned shorts, bright eyes, looking at the camera, leaning against a white post, with a blurred plant background, dynamic, alluring, light, energetic, shot with Canon R5, 70–200mm, cinematic studio lighting, 8k UHD, photographed by Bert Stern.’
The outcomes of this prompt are as follows:
The similarities between the original image and the .81 setting are striking! The details in the outfit's pattern are remarkably similar. Additionally, the prompt included the phrase "looking at the camera," and Leonardo AI responded adeptly. Once again, the lighting has been refined across all images, resulting in a more lifelike appearance than the original photograph.
By selecting an image with intricate details in the outfit, you can observe how Leonardo AI interprets the characteristics of the photo in relation to the strength of Image Guidance. Adjusting this strength can yield fresh, innovative imagery while preserving much of the realism from the original photo. Leonardo AI is both free and reasonably priced for its foundational model, making it accessible for experimentation.
Conclusion: The Future of AI in Photography
In the fast-paced world of AI, a week can feel like a year in other tech sectors. While Leonardo AI may currently hold an advantage over Midjourney, this could change rapidly. Nevertheless, the Image Guidance feature is a vital addition to any photographer's toolkit. It raises ethical questions for models, as they could potentially be replaced after one successful photoshoot, enabling photographers to create multiple new "shoots" without needing to rehire them.
Though the technology has not yet reached the realm of video, this is part of the reason why actors in SAG and Hollywood protested and successfully secured rights to their likeness. It makes one ponder whether a similar movement will arise within the modeling industry. AI continues to evolve, challenging the boundaries of ethics and creativity. For now, I will keep learning and making informed decisions regarding the use of various features in the future.
For further reading, consider these articles to deepen your understanding of AI:
- In Search of AI Perfection: Midjourney AND Leonardo.ai
- How to Add Realism, Personality, and Life to Midjourney AI Portraits
Make sure to subscribe to my email list to receive updates on my latest articles as soon as they are published.