Leonardo’s Pro Upscaler: Built for AI-Generated Images

While creators are increasingly working with AI-generated images, most upscalers are still trained on real photos only.

This mismatch creates constant friction. Standard upscalers often fail to read synthetic noise or stylistic grain correctly, twisting those intentional details into broken distortions rather than enhancing them.

We decided to solve this exact problem. Meet Leonardo’s Pro Upscaler: an upscaling tool designed specifically for repairing and enhancing AI-generated images.

Leonardo's Pro Upscaler

At Leonardo, we see the full workflow, from generating an AI image to editing and upscaling it. This perspective allowed us to understand the unique fingerprints of AI-generated images and build an upscaler that closes the loop.

Engineered to Recognize AI Textures

Most of the industry relies on upscaling models trained exclusively on real photographs. Because of this, when traditional upscalers process AI-generated images, they misinterpret the unique textures of AI content, introducing distortions that can make an image unusable.

To fix this, we engineered a custom pipeline specifically for AI-generated content. Leonardo’s Pro Upscaler understands synthetic grain and AI textures, cleaning them up rather than amplifying them. Your image scales perfectly without degrading.

The result is an output that maintains its strict structural integrity under intense magnification. In the example below, watch how the microscopic details remain sharp and coherent as we zoom in on the final upscaled file:

A True 105 MP for Any Digital and Print Project

An inflated 1,000-megapixel (MP) image filled with blurry textures and distortions is useless for a billboard. Leonardo’s Pro Upscaler delivers a true 105 MP that easily outperforms any 4K upscaler.

In the real world, a true 105MP output covers almost every professional use case. For digital work, it handles everything from social media thumbnails to large-scale 8K displays and digital billboards with total clarity:

For print, 105MP easily covers high-end retail displays, full-page magazine spreads, and A1 posters at a perfect 300 DPI. You can even push it to massive outdoor billboards with room to spare:

The dimensions in the table are modeled on a 3:2 image.

Precise and Creative Modes

Sometimes you just need to make a perfect image larger, and other times you need to repair structural issues while scaling. That is why we split the Pro Upscaler into two distinct modes: Precise and Creative.

The Precise Mode

Precise mode takes your existing image and scales it up without altering the core composition or adding new generative elements. This is the right choice when your initial generation is highly accurate and identity fidelity is non-negotiable.

A special feature of Precise mode is the Fix AI Image Artifacts toggle. AI artifacts are distortions that become more visible when you upscale. When you activate Fix AI Image Artifacts, our pipeline applies a custom algorithm that is specifically engineered to handle these AI artifacts.

For instance, if your original AI-generated image emulates a film look with heavy grain, upscaling it without Fix AI Image Artifacts active can amplify those imperfections:

Here’s how to choose your settings:

• Precise (Fix AI Image Artifacts off): Use this for images that are already clean and highly detailed, such as real photos or outputs from Nano Banana Pro. It acts as a flawless magnifier.
• Precise (Fix AI Image Artifacts on): Use this when your source image has synthetic noise, fake film grain, or minor compression artifacts. It’s highly recommended for outputs from Seedream or GPT Image.

The Creative Mode

Sometimes an image has structural problems that go beyond simple noise. If your generation has malformed hands, distorted background faces, or unresolved details, scaling it up will only make those errors larger. Creative mode fixes this by actively repairing and rebuilding content during the upscale process.

You can control the intensity of this generative reconstruction using three specific settings:

• Subtle: Applies a light generative pass. It gently refines edges and micro-details while keeping the original image strictly intact.
• Balanced: The best starting point for most problematic images. It actively repairs minor structural anomalies like background faces or distorted hands while maintaining the overall composition.
• Strong: Takes maximum creative liberty. Use this when your base image has significant structural errors and you need the AI to aggressively reinvent and reconstruct missing details.

To better understand the differences among these Creative modes, consider the example below, where the model wears an intricate haute couture gown covered in thousands of hand-stitched beads and metallic embroidery. The original AI generation has many unresolved details that become increasingly obvious as we zoom in. If we focus on the central red crystal on the model’s neck, we see that Creative Strong delivers the sharpest output, capturing the geometry and sharp light reflections of the embellishment better than the Subtle or Balanced modes:

How to Use the Pro Upscaler

On Leonardo, you can use the Pro Upscaler for both real photos and AI-generated content. Let’s explore each workflow.

AI-Generated Images

The quickest way to upscale an AI-generated image is directly from your feed. Hover your mouse over the image and click the Upscale image button:

A modal will appear where you can select your specific upscaling parameters. Once you choose your settings, click Upscale:

You can also run this process directly from the Upscaler page:

Click either the central Add image button or the New button in the top-left corner, then select the image you want to process:

Choose your upscaling parameters from the right-hand menu. When you are ready, click Upscale:

Pro tip: You can instantly transfer your upscaled image into Canva. This makes it easy to add typography for a billboard or apply any other final layout edits. To do this, click the Canva icon under the Upscaling menu:

This is how our billboard could look in Canva:

Uploaded Images

You can also upscale real photographs or AI-generated images created on other platforms. To start, navigate to the Upscaler page and click either the central Add image button or the New button in the top-left corner:

Bring your file into the system by clicking the Upload an image button:

Set your upscaling parameters using the menu on the right. For real photos, we recommend using the Precise mode with the Fix AI Artifacts toggle off. Once your settings are dialed in, click Upscale:

How to Choose the Upscale Multiplier

The Pro Upscaler offers an enlargement range between 2x and 8x, giving you precise control over your final image size. The AI does not apply the exact same processing across the board; it takes far more creative liberty at 8x than it does at 2x. That’s why it’s important to match your selection to your specific needs:

• 2x multiplier: Ideal for quick digital enhancements, such as social media posts or web banners. At this lower scale, the difference between Precise and Creative remains minimal.
• 3x to 6x multipliers: The sweet spot for high-resolution digital displays or standard print formats. Here, the AI actively repairs artifacts, and the outputs of the different modes begin to diverge significantly.
• 8x multiplier: Built for heavy-duty production work, such as large-format printing or massive billboard campaigns. At maximum scale, the upscaler takes progressively more freedom to reconstruct unclear or degraded details.

Pro tip: Don’t always pick the 8x multiplier just because it’s the biggest number. If your goal is a web banner, a 2x multiplier is enough, and choosing between Precise and Creative makes little difference. If your goal is a large physical print, you will need an 8x multiplier. At this extreme scale, the AI actively rebuilds the image, meaning your choice of Precise or Creative mode will significantly change the result.

A New Architecture for the Generative AI Era

Standard upscalers are held back by legacy architecture, relying on a philosophy built entirely for the pre-generative AI era. We realized this approach is obsolete and built an upscaler specifically engineered to recognize and handle the unique fingerprints of AI-generated images.

You can now execute your entire workflow directly on Leonardo, from generating your initial image to editing the fine details and scaling it up for production using the Pro Upscaler. We can’t wait to see what you create!

FAQs

What if my image is a real photo, not AI-generated?

The Pro Upscaler also works on traditional photography. For real photos, we recommend using the Precise mode with the Fix AI Artifacts toggle off. This configuration acts as a flawless magnifier, expanding your resolution while strictly preserving the integrity and reality of your original shot without introducing any generative alterations.

What does 105MP actually get me in the real world?

105MP gives you enough resolution for virtually any professional use case, whether it's print or digital. For print, 105MP is more than enough for billboard and large-format advertising, full-page magazine spreads, fine art prints, and high-end retail displays, with room to spare. For digital applications, the extra resolution gives you cropping flexibility. You can reframe, zoom in, or adapt an image for multiple formats without hitting a quality ceiling. That makes it useful for campaign assets that need to work across different placements and aspect ratios.

What is the difference between MP and 4K?

Both are ways of describing image resolution. "K" refers to the number of pixels along the longest edge of an image, so a 4K image has roughly 4,000 pixels on its longest side. Megapixels (MP) refers to the total number of pixels across the entire image, measured in millions. You find the MP count by multiplying the width by the height and dividing by one million. "K" gained popularity with the rise of digital video, where it's assumed that the aspect ratio is a fixed 16:9. Because still images have much more varied aspect ratios, "K" isn't a reliable headline figure for photos unless an exact aspect ratio is specified.

Can I run an image through the upscaler multiple times?

Yes, but that depends on your starting resolution. The upscaler works by multiplying your pixel dimensions, and the absolute output cap is always 105MP regardless of how many times you upscale. If you have a very small starting file, you might be able to process it multiple times before hitting the ceiling. However, a single 2x upscale on a 26MP (5,100 x 5,100) image takes you directly to the 105MP (10,200 x 10,200) limit because upscaling multipliers apply to both the width and height of your image.