Image-based prompting, blending and style control

Midjourney’s ability to incorporate visual references alongside text prompts unlocks a powerful creative toolkit. Whether you’re guiding an image’s structure using a reference photo, applying a consistent artistic style, or maintaining a character’s visual identity across multiple generations, image-based prompting offers precise control over the AI’s output.

In this section, we’ll explore how to use image prompts effectively, how they differ from image blending, and how key parameters like --iw, --sref, and --cref function across both the Web and Discord interfaces.

How to use image prompts and image blending

Although both image prompting and image blending involve using multiple images as input, they serve different creative purposes.

• Use image prompts when you want to direct the subject, layout, or mood of an image using visual references—paired with or without descriptive text.
• Use blending when you want to merge the visual elements of multiple images into a single composition quickly, without needing precise styling or identity control.

Note: On the Web UI, if you upload multiple image prompts without any accompanying text, Midjourney will automatically generate a blend-style output—similar to the /blend command in Discord.

Image prompt on Web UI

The Midjourney Web Interface now supports advanced image-based prompting through an intuitive, visual workflow. When creating a new prompt, you’ll find three dedicated reference sections:

• Image Prompts – Used to shape the overall content and layout of the output using one or more input images.
• Style References – Used to apply the tone, color palette, and visual treatment of an image (equivalent to --sref).
• Character References – Used to maintain consistency in character features like facial structure, clothing, or hairstyle (equivalent to --cref).

Note: The first section labeled “Frame” is used for video generation, which will be covered in the next chapter.

To use image prompts, you can drag and drop images directly from your device or gallery into each reference field. Then, add a short descriptive prompt.

For example, use this dog character image for image prompt and style reference.

Use the same dog character image as both the image prompt and style reference. Then enter “a cat” in the text prompt field.

Midjourney will generate a cat image that mimics the structure and style of the original dog image—preserving layout and brushwork characteristics.

For more in-depth examples, refer to the case study Testing Style and Character Reference Impact, where we compare the outputs of image prompt, style reference, and character reference individually.

[Box Yellow Start]

Blending images using multiple uploads

Using image prompts is simple and flexible. When you upload multiple images without any text, Midjourney automatically blends the visual elements into a single unified result.

For example, upload the following two images into the image prompt section.

Then hit Enter without typing any text.

Midjourney will generate an integrated image that merges key visual features from both inputs.

This technique is a quick way to explore combinations of design concepts, character poses, or styles—even without writing a single word.

[Box Yellow e]

Refining image prompt with image weight, style reference, and character reference in Web UI

In the Midjourney Web UI, you can fine-tune how each image contributes to the final result by assigning it a specific role: image prompt, style reference, or character reference. Each plays a distinct role in shaping the composition, visual tone, and subject identity.

The interface provides a structured layout where you can upload images into clearly labeled sections. While assigning roles is visual and intuitive, adjusting the strength of each reference still requires adding text-based parameters like --iw, --sw, and --cw to your prompt.

Image prompt and --iw

The image prompt acts as the primary reference. It influences the layout, pose, color composition, and subject matter of the output. You can control how strongly this image affects the result using the --iw (image weight) parameter.

• Default: 1.0
• Range: 0.5 to 3.0
• Affects: Only the image prompt (not style or character references)
  
  

Style reference and --sw

The style reference determines the artistic qualities of the result—such as brush strokes, lighting direction, and color palette. It’s useful when you want to retain the content structure of one image but express it with the visual mood of another.

• Optional modifier: --sw (style weight) controls the influence of the style reference
• Range: 0 to 1000
• Default: 100

Character reference and --cw

Character references help maintain visual identity—such as facial features, hairstyle, or clothing—across different scenes. This is especially useful for storytelling, branding, or maintaining continuity in a character’s appearance.

• Optional modifier: --cw (character weight) controls how closely the AI preserves the character’s likeness
• Range: 0 (very loose) to 100 (very strict)
• Default: 100

Case Example: Adjusting influence with prompt weights

To better understand how each reference type and parameter affects the final output, let’s look at a practical example using the Midjourney Web UI.

We used three input images, each assigned a different reference role in the Web UI:

Image Prompt: Defines the structure and subject framing

 

Style Reference: Influences color palette and overall mood

 

Character Reference: Aims to preserve facial and character identity

Case 1: Default influence

In this case, we used a simple text prompt:

Even though we didn’t explicitly apply any cartoon reference for style, the AI leaned into a stylized, cartoon-like interpretation. This suggests that all references were balanced by their default weights, allowing a playful blend of structure, mood, and identity.

Case 2: Adjusted weights for stronger control

For this version, we applied additional parameters to strengthen the structural and stylistic influence, while relaxing the character likeness:

Here is the output. You can see that the image retains the essence of the original while incorporating influences from the additional reference images.

Even with the same set of reference images and text, small changes in --iw, --sw, and --cw can shift the balance between content, style, and character. Use default values for inspiration and experimentation. Add weights when you want to sharpen direction and focus.

Image prompt in Discord

In Discord, you can guide image generation by pasting one or more image URLs directly into the /imagine prompt, followed by a text description. These images act as visual anchors that influence the structure, composition, or subject of the output. To control how strongly an image prompt affects the result, use the --iw (image weight) parameter.

How to get an image URL in Discord

1. Upload an image into any Discord channel (including your own Midjourney thread).
2. Right-click (or long-press on mobile) the uploaded image.
3. Select “Copy Link” – this will give you a direct image URL.

The copied URL will look like this:

https://cdn.discordapp.com/attachments/123456789/sample-image.jpg

You can paste this URL directly into your prompt.

This setup tells Midjourney to use the uploaded image as a reference, while also considering the text “a futuristic vehicle design.” The --iw 2 makes the image twice as influential as the default.

Style reference with --sref and --sw

To apply the style (lighting, brushwork, color palette) of another image:

• --sref: Specifies the style reference image.
• --sw: Style weight.

Character reference with --cref and --cw

To maintain character identity from a reference image:

• --cref: Specifies the character reference image.
• --cw: Character weight.
  
  

Combining all in one prompt

You can use image prompt, style reference, and character reference together:

This command will:

• Use pose.jpg for composition and layout,
• Apply the visual tone of style.jpg,
• Keep the character identity from char.jpg.
  
  

Using Discord in this way gives you full control over structure, style, and subject identity—directly from a single prompt.

/blend command in Discord

The /blend command offers a simplified way to merge up to five images without requiring a text prompt or parameter adjustments. It’s designed for fast and intuitive visual experimentation, ideal when you want to generate hybrid concepts or explore stylistic fusions.

To use it:

1. Type /blend in any prompt channel.
2. Upload 2 to 5 images in the provided fields.
3. Choose an aspect ratio from the dropdown: (1:1 (square) is the default, 2:3 for a portrait-style output, and 3:2 for landscape-style output)
4. Submit the command to generate a blended result.

Unlike image prompts used with /imagine, the /blend command does not support text input or parameter adjustments such as image weighting. Instead, it automatically fuses the visual characteristics of the uploaded images into a single cohesive result. It’s ideal for quickly exploring visual combinations without needing to fine-tune prompt behavior.

You can explore blending example case studies demonstrating blending techniques below:

• Blending Dragon, Robot, and Pyramid Worlds
• Blending One Character into Different Worlds

---

Image-based prompting in Midjourney unlocks a deeper level of creative control—allowing you to shape structure, apply visual styles, and maintain character identity with precision. Whether you’re guiding the composition with an image prompt, styling the output with a reference artwork, or ensuring consistency with a character likeness, these tools give you the flexibility to realize highly customized results.

While both Discord and the Web UI offer powerful workflows, the Midjourney Web UI now provides a more streamlined and intuitive experience. With clearly labeled input areas for image, style, and character references, it removes much of the guesswork from prompt construction. And despite requiring text-based parameters like --iw, --sw, and --cw for advanced control, the visual interface makes the setup far more approachable—especially for those new to Midjourney or working across multiple images.

For fast blending, the /blend command in Discord remains a useful option. But for more nuanced creative direction, the Web UI is now the preferred environment—bringing clarity, ease of use, and enhanced functionality to the image creation process.

As you continue experimenting, revisit your image prompt strategies often. Even subtle changes in image role or weight can dramatically shift the output—and that’s where the true creative power lies.