How to Use Midjourney's Style Tuner to Save and Reuse Visual Styles
The Style Tuner is one of Midjourney's most practical tools for creators who want visual consistency across a series of images. It works differently from --sref (style reference) in that it does not copy a specific image's style but instead lets you define a custom style direction by making a series of visual preference selections. The result is a style code you can reuse across any number of prompts to pull all outputs in the same aesthetic direction.
This guide covers how to run the tuner, what the selection process actually does, how to apply and adjust the resulting code, and when Style Tuner is more useful than the --sref approach.
How the Style Tuner Works
The Style Tuner generates a set of image pair comparisons from a base prompt you provide. For each pair, you choose which image better represents the visual direction you want. After making your selections across all pairs, Midjourney generates a unique style code that encodes your preferences as a parameter you can append to any future prompt.
The style code captures a directional aesthetic rather than copying a specific image. It is a vector in Midjourney's style space that reflects your selections, biasing future outputs toward the visual characteristics you chose across the comparison pairs.
Starting a Style Tuner session
Use the /tune command in Midjourney:
/tune [your base prompt]
Midjourney asks you to choose:
- Style directions: 16 or 32 image pairs to compare. More pairs give more granular style definition at the cost of more comparison choices to make.
- Aesthetic mode: Default or Raw. Default mode applies Midjourney's aesthetic enhancement to all generated pairs. Raw mode uses a more neutral baseline closer to the literal prompt. For style building, Raw is often more useful because the aesthetic choices you see are less filtered by Midjourney's own style preferences.
After confirming, Midjourney generates all the comparison pairs and sends you a link to the Style Tuner interface.
Making Style Selections
The tuner interface shows a series of image pairs side by side. For each pair, you can:
- Select the left image
- Select the right image
- Select neither (if neither image represents what you want)
The quality of your final style code depends almost entirely on how consistent and intentional you are during the selection phase. Several principles help:
Decide what you are optimizing for before you start
Are you selecting for color palette, rendering style, level of detail, mood, or compositional approach? Mixing criteria produces incoherent style codes. Decide on one or two primary dimensions before the session and evaluate every pair against those dimensions only. If you are building a moody, dark atmospheric style, reject any warm or light-toned image regardless of other qualities. Consistency in the selection logic is what makes the resulting style transferable.
Use "neither" actively
The neither option is not a cop-out. If both images in a pair represent a direction you actively do not want, selecting neither is more useful than picking the slightly less bad option. An unclear signal is worse than no signal.
Work quickly and intuitively on first pass
Initial instinctive responses to the pairs are often more reliable than overthought decisions. Going back and changing many selections after the fact tends to produce less coherent codes because the reasoning becomes inconsistent across pairs.
Applying Your Style Code
After completing the selections, Midjourney provides a style code that looks like a string of characters. You apply it to any prompt using the --style parameter:
your new prompt --style [your-style-code]
You can also combine the style code with --stylize to control how strongly the style is applied:
- Low --stylize values (50-100): the prompt description has more influence, the style has less
- High --stylize values (500-1000): the style code has more influence, the prompt has less
- Default --stylize (100) is a balanced starting point
The style code is persistent. It does not expire and can be shared with others. If you are working on a series and want collaborators to match your visual direction, sharing the style code lets them apply the same style without needing to match your exact prompt wording.
