Midjourney drops Style Tuner to let you lock in custom visuals

Anyone who has spent time with Midjourney knows the frustration of prompt lottery. You can paste paragraphs of descriptors, but getting consistent lighting, texture, and mood across multiple renders remains an uphill battle.

Midjourney rolled out a new feature called Style Tuner, designed to let users visually tune and bottle their own aesthetic into reusable codes.

Pick your visual vibe

Using the tool is straightforward and requires no separate registration.

It all starts with a new Discord command: type /tune followed by your subject, like /tune A curious cat. You then choose how many base style variations you want Midjourney to generate. The cost in GPU time scales with the number of styles you request, and Midjourney displays the estimated GPU time upfront before you hit submit.

Once the generations finish, Midjourney sends a direct message containing a custom link pointing to tuner.midjourney.com.

Opening that URL takes you to an interactive page showing side-by-side style pairs. You click your favorites—guides suggest picking around 5 to 10 pairs—and the tool compiles your choices into a short alphanumeric style code.

From there, you just append --style CODE to any regular /imagine prompt to lock in that aesthetic.

Stacking codes and pulling wildcards

The feature also introduces several flexible workarounds.

If you do not want to spend GPU hours generating a tuner page, you can type --style random to roll the dice on unexpected aesthetics.

You can also chain multiple styles together using a hyphen, such as --style code1-code2, effectively blending two distinct visual flavors into a single prompt.

Meanwhile, the existing --stylize parameter controls how strongly the custom style influences the final generation. You can also view the tuner matrix behind any existing code by appending it to the tuner URL.

One caveat is that styles are tuned to the initial prompt. While a cat-derived style will likely hold up on a dog portrait, it might behave unpredictably when applied to architecture or sci-fi landscapes.

Why it matters

Style Tuner tackles the single biggest friction point in generative art: visual consistency.

Up until now, locking down a signature look meant hoarding dense artist names and camera specs that broke whenever the underlying model shifted. Packaging aesthetic taste into portable, shareable codes gives creative agencies and solo designers a repeatable playbook.

We think shifting the tuning process out of Discord into an interactive web selector is brilliant. Instead of guessing prompt syntax, users simply make taste decisions.

The downside is the steep GPU cost required to generate the initial tuner grids, which might deter casual hobbyists from experimenting freely. Still, turning aesthetics into shareable text parameters creates an entire ecosystem of user-curated styles that competitors will have to answer.