Midjourney New Features Explained: HD Upscaling, Moodboards, and Personalized Models

Midjourney’s recent rounds of updates have filled in the gaps in editing and style management beyond just “image generation”: HD upscaling no longer relies on third-party tools, moodboards make styles more controllable, and personalized models make it easier to produce results that match your aesthetic. Below, I’ll break things down by feature and also share the most hassle-free way to use them in Discord.

HD Upscaling: Sharper enlargements, more stable details

In the past, many people would take images generated by Midjourney to external tools for a second round of upscaling. Now Midjourney’s own HD upscaling experience is more complete—you can upscale directly after the image is generated. The practical benefits are cleaner edges and more natural textures, making it suitable for detail-sensitive scenarios such as e-commerce hero images, poster text borders, and human skin.

It’s recommended to generate images using the default workflow first, then upscale the one you’re most satisfied with. If you find the details “drift” after upscaling, it’s usually more time-efficient to go back to that same image, create variations, and then upscale, rather than rewriting the prompt.

Moodboards: Lock in a brand style

Midjourney’s Moodboards are more like a “style binder”: you upload a set of reference images, and the system extracts shared color tones, materials, and overall visual vibe, making subsequent generations easier to keep consistent. This is very friendly for teams that need long-term, stable output—such as brand key visuals, app illustrations, and short-video covers.

When using them, don’t dump in a random pile of images. It’s better to include only a single visual language (consistent palette, camera approach, lighting and shadow) within the same moodboard, so Midjourney can more easily learn the specific “feel” you’re aiming for.

Personalized Models: Make Midjourney more like “your aesthetic assistant”

In this update, Midjourney has upgraded personalized models. The key change is a stronger emphasis on “preference consistency”: the more often you choose the same type of imagery, the more likely it is to generate results that match your taste. It doesn’t write prompts for you; instead, it turns the styles you repeatedly select into default tendencies.

If you’re creating a serialized character or multiple scene images within the same project, it’s recommended to stick to a fixed prompt structure and use the personalized model to stabilize the style, rather than starting from scratch for every image.

Model and Command Details: Control results better with /settings and --version

Midjourney still allows switching model versions using --version (or --v), and you can also type /settings in Discord and select a model in the panel. Different models have different strengths in photorealism, illustration, and material detail. If your outputs feel unstable, first check whether the model version matches the style you want.

If you want to reduce the “gacha” feeling, one simple approach is: for the same project, keep the model version, aspect ratio, and core descriptive keywords as consistent as possible, so Midjourney can iterate on details under stable conditions.