Midjourney V8, Explained Simply

Midjourney V8 is a newly released, early-test version of the popular AI image generator. It is perfect for artists, creators, or anyone who wants to turn very specific, detailed ideas into high-quality pictures, especially if you want the AI to match your unique art style.

What it does well

• Follows directions: It is much better at listening to exactly what you ask for in your prompts, especially if you write longer, detailed descriptions.
• Understands your style: It is amazing at using "personalization" (teaching the AI your preferred look), style references, and moodboards to create exactly the vibe you want.
• Clearer images and text: The pictures are more detailed and make more sense visually. If you put words in “quotes,” it can actually spell text correctly.
• Speed: It creates images about five times faster than the previous model.
• New interface: The website has a new conversation mode so you can chat naturally, a "Grid Mode" for looking at big sets of images, and sidebars so you can change settings without blocking your screen.

How it compares

• Versus Midjourney's older model: V8 is a completely new system. It is much faster and better at following complex instructions than V7. However, the "default" look of V8 is still being fine-tuned, so it might require totally new ways of writing prompts compared to what you are used to.
• Backwards compatibility: Even though it is a new model, it still safely works with the personalization profiles, moodboards, and style references you might have already saved in V7.

What it costs and what it can handle

• Speed and Cost: Regular image generation is very fast. However, special features like --hd (which natively renders images at a sharp 2K resolution), --q 4 (a mode for extra visual coherence), style references, and moodboards currently cost four times more and take four times longer to generate.
• Supported formats: It supports multiple aspect ratios (different image shapes like wide or tall) and special commands like --chaos, --weird, --exp, and --raw (a mode for a more realistic, plain, "in-control" photo look).

Worth knowing

• No Relax mode yet: "Relax" mode (which usually lets paid users make images slower for free) is not supported yet. The team is working on new servers to bring it back and make rendering cheaper.
• You have to guide it: Because the default aesthetic is still being worked on, you shouldn't just rely on the basic look. The makers highly recommend turning on your personalization settings, cranking the --stylize setting up to 1000, or using --raw for photos.
• It is an Alpha test: This is an early "Alpha" release, meaning it has unique strengths and weaknesses. The creators actively want you to rate the images you generate (using the "I like this" or "I don't like this" buttons) so the AI can learn and improve.