Creating Better Midjourney Images: Key Tips and Techniques

Midjourney is a powerful tool for creating images with AI. To go beyond basic results, you need to understand its features well. This guide covers essential tips, from using parameters to leveraging advanced features like reference images and personalized styles, based on insights from an experienced concept artist's workshop.

Getting Started with Midjourney: Web vs. Discord

When Midjourney first appeared, it used Discord. Now, the company's focus is shifting towards the web interface. The web UI offers almost all the settings you find in Discord, and using it is the recommended way forward. Midjourney updates its model and features often, learning from what users like. Using the web interface helps the company gather user preferences faster, guiding future updates.

Midjourney's official website is a key resource. It has a detailed manual that is updated weekly. If you want to truly understand and use Midjourney effectively, always check the official documentation.

Understanding the Midjourney Prompt Structure

A Midjourney prompt has two main parts: the text description and parameters. While you can get images with just text, parameters give you more control. The most impactful element in your prompt is often an image prompt, which acts like a reference image. Telling Midjourney "I want something like THIS image" strongly guides the result.

After the image prompt (if used), the text prompt is next in importance. Midjourney's model is unique and is refined regularly by the company.

Weight is another important factor. You can assign weight to different parts of your prompt or to reference images, significantly changing the outcome. Midjourney's model also gives importance to style. Putting the desired style first in your text prompt can help achieve the look you want. A good order for your text prompt is: style, situation, description, and atmosphere.

Essential Midjourney Settings and Parameters

Beyond the basic prompt, several settings let you fine-tune your images:

Stylize (`--s`)

This setting controls how much Midjourney's default aesthetic is applied. It ranges from 0 to 1000, with 100 as the default.

• Higher stylize values (closer to 1000) make Midjourney decorate the image more, often making it look more artistic or "Midjourney-like."

• Lower stylize values (closer to 0) make Midjourney stick closely to your prompt words, resulting in a less decorated image.

If you want images with more variety, increase stylize. If you want to generate the theme you described precisely, keep it lower. For those looking to streamline their experimentation with different settings and parameters, consider using the TitanXT Midjourney Automation Suite. It can help manage variations and track results.

Style Raw (`--style raw`)

If you want a very realistic image, like a photograph, check the Style Raw option.

Aspect Ratio (`--ar`)

Midjourney is very flexible with aspect ratios, unlike some other AI image generators that might distort images if the ratio is too different from their training data. You can set the width and height ratio freely without getting strange results like extra limbs or faces.

Chaos (`--c`)

Midjourney usually generates four initial images from a prompt. The Chaos setting impacts how different these four images are from each other. It ranges from 0 to 100.

• Low Chaos (e.g., 10) results in four images that are quite similar.

• High Chaos (e.g., 100) results in four images that are very different, sometimes adding elements not explicitly in the prompt but related to the core idea.

The "Variety" option or tab often relates directly to the Chaos setting.

Weird (`--w`)

Increasing the Weird setting makes the generated images look more unusual.

No (`--no`)

This parameter tells Midjourney what to exclude from the image. If your prompt generates something you don't want (like a person appearing behind a main subject), you can add `--no men` to try and remove it.

Permutation Prompts (`{a,b,c}`)

This feature lets you generate variations of a prompt quickly. By putting options in curly braces, you can tell Midjourney to create images for each option or combination. For example: `A {red, blue} car` would generate images of a red car and a blue car. This saves time when exploring different concepts or variations like aspect ratios or keywords. Keep in mind that permutation prompts use processing time, which can impact your 'faster hours' usage if you are on a plan that charges for it.

Advanced Midjourney Features: Personalization and Reference

To truly make Midjourney work for your specific vision, explore these advanced tools:

Image Ranking and Personalized Code (`--p`)

Midjourney allows users to participate in image ranking. By comparing two images provided by Midjourney and selecting which one you prefer, you give the system data about your taste. After ranking over 200 images, Midjourney gives you a personalized code (`--p`). This code represents your unique style preferences. Using this code in your prompts helps Midjourney generate images that align with your personal aesthetic, making your results less generic.

You can have separate personalized codes for the standard Midjourney model (realistic) and the Niji Journey model (anime/cartoon style). These codes significantly impact the style and tone of your generated images.

Style Reference Seed (`--sref`)

Separate from your personal style code, there's a style reference seed. These are specific numbers assigned to various stylistic aesthetics within Midjourney (there are billions!). While Midjourney gives you a personalized code based on your ranking, you need to find style reference seeds. Many websites collect and share these seeds with example images. If you find a style you like, you can include its seed number (`--sref [number]`) in your prompt to apply that style to your image. You can even combine your personalized code and style reference seeds for a highly customized result. Trying out different combinations to find a unique style can be time-consuming. Tools like the Midjourney Automation Suite from TitanXT could potentially help manage the process of testing various seeds and codes.

Reference Images (Image, Character, Style)

Uploading your own images allows you to use them as references in three main ways, each with a different impact and weight parameter:

• Image Reference (`--iref`): Weight 1-3. Uses the overall composition, colors, and feeling of the reference image.

• Character Reference (`--cref`): Weight 0-100. Focuses on copying the features of a character from the reference image. Lower weights (0-1) might only copy facial structure, while higher weights (50, 100) pull in hairstyle and other overall features.

• Style Reference (`--sref`): Weight 0-1000. Applies the artistic style or color palette of the reference image to your prompt. This is different from the style reference *seed* but achieves a similar goal by demonstrating a style through an image.

You can mix these references. For example, use an image reference for the scene, a character reference for a subject, and a style reference image or seed for the aesthetic. This layering of references allows for complex and specific results. You can also use the Shift key when clicking reference images in the UI to apply multiple references at once.

Midjourney Inpainting and Editing

Midjourney allows you to edit parts of your generated images. The Inpainting feature lets you mask an area and then prompt for changes within that masked region. You can use reference images or text prompts to guide the edit. For example, you could mask a character's face and use a different image as a character reference to paste that face onto the body.

Editing features like changing size or view are also available. While Midjourney's text application within images has improved (especially after version 6), it's still not perfect. Think of AI tools not just for creating from scratch, but also for editing and refining existing work quickly.

Remix Subtle

The Remix feature lets you change the prompt or add references when varying an image. Remix Subtle is a setting that keeps the original image very similar while applying only small, detailed changes based on your new prompt or reference. This is useful for making minor adjustments without losing the core image.

Retexture (Experimental)

An experimental feature called Retexture (found in the Edit menu) allows you to apply the texture, silhouette, or atmosphere from a reference image onto another generated image. This feature is currently limited to experienced users, but as it becomes more stable, Midjourney often makes these features available to all users within a few months.

Beyond Still Images: AI Video Generation Concepts

While primarily known for still images, Midjourney and similar AI models based on "transformer architecture" are relevant to video. Models good at 'linking' images can predict the next frame, which is key for video. There's a crucial difference in mindset for image vs. video prompts:

• Image Prompts: Setting. You describe a scene, atmosphere, and elements, like setting up a stage.

• Video Prompts: Directing. You need to describe movement and changes over time, like directing action within the scene.

A common workflow for AI video is to generate key scenes or images using Midjourney, then use AI video tools to make them move. Combine this with AI-generated audio (voice, music, effects) and edit everything together using video software. Video enhancement tools can improve the final quality for real-world applications.

Conclusion

Midjourney is more than just a simple text-to-image tool. Achieving specific, high-quality results requires understanding its various parameters, reference options, personalized codes, and experimental features. The core idea is that you need to guide the AI using these technical details, not just basic descriptions.

By learning how to use image, character, and style references, adjusting weights like stylized and chaos, using the 'no' parameter, and leveraging personalized codes, you can move beyond stereotypical AI art and create images that truly match your creative vision. As AI tools continue to evolve, combining different techniques and understanding the underlying concepts will be crucial for creators. Exploring tools that help manage and automate these complex workflows, such as the TitanXT Midjourney Automation Suite, can further enhance your ability to experiment and produce refined results efficiently.