Blending images: creating new visuals through image fusion

Blending images allows you to merge distinct visual concepts into a single frame—whether it’s fusing natural and cosmic landscapes, or combining fantasy with realism. It’s a technique that opens doors to imaginative storytelling through visuals.

The blending feature and its output vary depending on the platform:

Midjourney supports image blending on both the web interface and Discord. The Web UI provides a simplified blending workflow, while Discord gives you more control by allowing prompt-based customization—such as assigning weight using multi-prompt syntax. However, detailed control over structure or positioning is still limited.

Leonardo AI offers a Canvas editor with AI-assisted inpainting that allows you to blend and recompose sections with higher precision. This gives users more authority over positioning and seamless transitions.

ChatGPT-4o enables image blending in a conversational flow. While there’s no support for numerical weight control or layers, you can still influence results through thoughtful instruction. The tool usually overlays images and merges style and texture holistically.

In this section, we’ll use ChatGPT-4o to demonstrate two common workflows: blending to create new scene compositions and blending to create new characters.

If ChatGPT isn’t working or refuses to process the task, switch to Sora. You can access Sora from the left sidebar. Sora provides a user interface focused on video and image generation. Click the + button to upload images. Choose the aspect ratio you want to use, then type your prompt in the message box.

Blending images to create imagery scenes

We’ll begin by blending entire scenes. The goal here is to combine visuals from different thematic elements—land, sea, and cosmos—into unified, imaginative environments.

There are the input images:

Natural mountain range

Bioluminescent coastal nightscape

Cosmic humanoid profiles

After uploading a pair of images, simply ask ChatGPT:

When blending in ChatGPT-4o or similar models, it’s often better to:

• Keep the prompt short and focused. Overly long prompts can distort how image features are layered.
• Use source images that are the same size and aspect ratio. This helps the AI align elements more predictably during overlay-based blending.

This minimal approach tends to produce more cohesive and balanced compositions in current-generation AI tools.

Here are the resulting blends:

Blended Image: Mountain + Coastal Nightscape

Blended Image: Mountain + Cosmic Profiles

Blended Image: Coastal Nightscape + Cosmic Profiles

Each blend showcases how distinct visual themes—natural light, galactic energy, and oceanic flow—can harmonize and even enhance one another when fused creatively.

Note: As with scene blends, it’s important to start a new thread or conversation for each blend. This helps avoid unintended influence from previous results or images, keeping the output clean and focused on the current input pair.

You can also experiment by blending all three images together for an even more layered and complex visual.

Blending all three images together

Blending images to create new characters

Using the blending technique, you can also generate entirely new characters by merging two different visual styles—one serving as the base (or seed) and the other as a stylistic influence. This approach is particularly useful for character transformation, redesigns, or applying a new aesthetic to an existing form.

In this demonstration, we combine a simplified 3D model of a hooded figure with a vibrant portrait-style illustration. The final result preserves the core structure of the seed while adopting the mood, color palette, and surface treatment of the style reference.

When blending for character creation, it’s best to use simple portrait-style images. Complex backgrounds or varied poses often confuse AI models during overlay processing. Centered, clearly defined characters produce much more coherent and controllable results.

Here are the two input images:

Seed character (base form)

Style reference (vibrant 2D portrait)

After uploading the two images, use the same minimal prompt:

This straightforward phrasing helps the AI focus on layering and stylistic fusion without introducing unrelated elements. It also works best when both images share similar orientation and framing.

Here is the result:

The final image retains the silhouette and pose of the seed character, while integrating the style reference’s fashion details, color intensity, and lighting. The blend feels intentional and stylized—a compelling example of how AI-driven image fusion can produce something new, imaginative, and visually cohesive.

Blending images opens up a wide space for experimentation and creative discovery. Whether you are merging photos to build something entirely new or simply enhancing your visuals with subtle overlays, this process invites you to see familiar elements in unexpected combinations. As you practice, you’ll develop a sharper sense of how composition, color, and style can work together to produce cohesive, compelling results.