New Midjourney V7 update brings faster rendering, more realism, and full sref compatibility

Midjourney V7 is now available, and the update focuses mainly on improving speed, image quality, and compatibility with existing workflows. While this isn’t a complete overhaul, it does address several things that can be useful for creators.

One of the biggest changes is speed. The company claims that V7 renders images about 20–30% faster than the previous version. Also, the difference should be more noticeable when you’re working with detailed prompts or trying to generate a series of images in one go. As per the update page, it holds up better during peak server times, which should make the whole process more consistent.

Image realism has also been bumped up. Skin, fabric, and lighting details should look a bit more natural. Human poses are less stiff, and expressions have more nuance, which helps avoid the “uncanny valley” effect (the uncomfortable feeling people get when something looks almost human but not quite, like a robot or digital face that seems slightly off) that sometimes showed up in earlier versions. If you're using prompts that include people, architecture, or natural landscapes, you should be able to see some differentiators.

The update also keeps full support for V6.1’s sref codes. That means users can continue using their saved styles without needing to rebuild them for the new version. The same references still work and might even look better with the upgraded model.

Using V7 is pretty simple: all users have to do is add "--v 7" to your prompt (minus the apostrophes). If you want to take advantage of the new personalization system, you’ll need to rate about 200 images at midjourney.com/rank-v7. After that, V7 uses your choices to generate more tailored results.

Other parameters like --ar for aspect ratio, --sref for style references, and --seed for randomization still work as before. So if you’ve been using Midjourney for a while, switching to V7 won’t require learning anything new.