Exploring Midjourney V6 Alpha: Text, Realism, and New Controls

A Midjourney generated image using Midjourney Automation Suite

Midjourney version 6 Alpha is now available for all users on Discord. This first release brings some big changes. We will explore three key areas: the new text feature, how image realism is better, and how two useful functions work.

Getting Started with V6 Alpha

Since version 6 Alpha is not the default yet, you need to turn it on. Go to your settings and pick the Midjourney model V6. Alternately, just add `--v6.0` to the end of any prompt you use.

Prompting works differently in version 6. Midjourney says we need to learn new ways to write prompts here. V6 understands prompts much better than before. It also reads more words in very long prompts. This means instructions you put at the end of a long prompt in V5.2, which might have been missed, are now seen in V6.

Because V6 is more accurate, Midjourney advises against using "junk tokens" in your prompts. These are words or phrases that don't add real creative direction, like "award-winning," "photorealistic," "render," "4K," or "8K." These words were often used in older versions to try and force a certain look, but V6 gets better results by focusing on clearer instructions.

Adding Text to Midjourney Art

Many users wanted the ability to add text to images. Midjourney version 6 now has a basic text feature. You must put your text inside quotation marks ("like this"). Midjourney suggests using `--style raw` or lower stylized values, like 100, to get better text results.

Seeing Text in Action

Let's look at some examples of the new text feature.

[LI]A prompt for a SpongeBob character holding a sign worked well.[/LI]

[LI]Rendering a blue hedgehog with a "let's ride" caption added nicely.[/LI]

[LI]A handwritten sticky note reminder to wash the car came out clearly.[/LI]

[LI]A birthday cake with a message looked good enough to eat. The colors were also nice.[/LI]

It seems that shorter text works better than trying to include long sentences.

[LI]Adding the word "bonjour" to a Paris cafe sign was successful. Using phrases like "single sign" or "single caption" in the prompt can help prevent text from showing up more than once.[/LI]

[LI]An old-fashioned diner restaurant with a neon sign rendered wonderfully. The sign looked fantastic.[/LI]

Creating many image variations with slight text changes can be time-consuming. Speed up your process and manage your projects more effectively using the Midjourney Automation Suite from TitanXT.

Improved Image Quality and Realism

You will notice more detail in V6 images, especially with people and other subjects. The results are more realistic.

Examples of New Realism

[LI]Close-up shots of a girl from the desert with her camel showed better detail in clothing material and patterns. Skin looked more real, not airbrushed. You could see finer strands of hair and eyelashes.[/LI]

[LI]Candid photos of couples on a boat ride captured the natural feel. Hair and skin details were sharp and believable. Facial expressions looked very natural, with realistic wrinkles around eyes and foreheads.[/LI]

[LI]In close-ups of a young woman doing an art project, her hands held the tools correctly. This was often a problem in V5.2 and other AI art programs. The AI also placed paint colors realistically smeared on her arms and hands.[/LI]

The level of detail has taken a step forward. While inpainting isn't in this Alpha yet, these examples show promise. Generate multiple high-quality, realistic images quickly. The TitanXT Midjourney Automation Suite can help you manage bulk image generation and variation testing efficiently.

Using Very Subtle and Very Strong Functions

Midjourney V6 has functions that allow you to make changes to a generated image. The "very subtle" and "very strong" functions let you control how much the change affects the image.

How Subtle and Strong Look

[LI]Using an image of an Angelina Jolie lookalike: Applying "very subtle" to change her jacket color and add sunglasses worked, but the background colors faded. Applying "very strong" with the same changes enhanced the colors everywhere, made her pose more certain, and darkened the sunglass lenses. The "very strong" result looked better.[/LI]

[LI]Trying to add Captain America's shield to a pose in front of a skyscraper: Using "very subtle" worked well to add just the shield. To make a bigger change, like putting the skyscraper on fire *and* adding the shield, "very strong" was needed. The results from "very strong" were impressive, with good color depth and accuracy.[/LI]

[LI]Testing a type of inpainting (changing one subject to another): Starting with a watercolor image of a moose by a river, using "very subtle" to change the moose to a deer resulted in a good-looking male deer. Using "very strong" for the same change kept the deer true to the request and enhanced the colors. This result was also very nice.[/LI]

Experimenting with the very subtle and very strong parameters can lead to interesting variations and control. Simplify your workflow for generating and comparing these types of variations. Explore the features of the TitanXT Midjourney Automation Suite to manage your creative process.

Current V6 Alpha Features

At the time of this video, version 6 Alpha includes the improved text drawing, better image coherence, and more accurate prompt following. Updates planned soon will bring improvements to drawing text, even better image coherence and prompt accuracy, and faster and cheaper upscaling.

Future updates will also look at how well version 6 responds to artistic styles, camera angles, and other photography details.

Version 6 Alpha is a positive step forward for Midjourney. It brings requested features and better image quality. Try it out and see what artworks you can create!