Midjourney v6 Alpha is Here!
How to use Midjourney v6
What's new in Midjourney v6 alpha?
Drawing Text with Midjourney v6
Improved Prompt Accuracy
Midjourney v6 Alpha Features
Summary
Just before the new year, Midjourney released a new model v6 alpha. As the name indicates, this is an alpha version of their new v6 model. Things will change (for the better) and and more features will be added. Here's a quick rundown of what you can expect with this new release.
How to use Midjourney v6
To start using Midjourney v6, you can type /settings in Discord and select v6 from the dropdown. This will set v6 alpha as the default model for your prompts. Alternatively, you can add --v 6 to the end of your prompts.
What's new in Midjourney v6 alpha?
According to Midjourney's announcement in Discord, v6 has better prompt accuracy, can handle longer prompts, has improved image coherence and more knowledge. It also has better image prompting and remix. Midjourney v6 can also draw text! I know a lot of people have been waiting for that. You'll also find improved 2X upscalers with both 'subtle' and 'creative' modes.
If you'd rather watch a video instead of reading this article, check out my YouTube video here!
Over the holidays I spent some time playing with v6 and I am especially excited about the text drawing capability and the improved prompt accuracy.
Drawing Text with Midjourney v6
To create images with text, include the text that you want drawn in quotes in your prompt. Midjourney will recognize that any text placed inside quotes should be drawn as text on the image.
In the example below, I put the text "FREE HUGS" in quotes within the prompt. Note that Midjourney will pay attention to capitalization.
Prompt: /imagine a cute, happy smiling raccoon standing in front of a boutique shop holding a beautiful sign that says "FREE HUGS". --ar 16:9 --v 6
I had a lot of fun creating a whole series of example images for my made-up sci-fi short-film called "Saturn is Closed".
I also made an entire winter holiday-themed alphabet. First, I had Midjourney make a letter "A". Then, once I got one I liked, I ran variations (clicked a "V button") on that and just changed the letter in the quotes to a different letter. This worked surprisingly well and it only struggled with a few letters: E, F, and Q.
Base prompt for a single letter: /imagine the capital letter "A", beautiful detail and cute Christmas creatures, Christmascore, cool-toned colors, dark blurry bokeh background
Drawing text with Midjourney v6 alpha is definitely not perfect, but it will get much better. Currently it works best for single words or short phrases. It also helps if the words are common and not obscure. Sentences are difficult to generate. Using a lower stylize value or adding --style raw to your prompt can sometimes improve the text results.
/imagine person wearing a T-shirt that says "hello it's me", full body --ar 3:4 --v 6
/imagine a rustic wood sign with the words "I became insane, with long intervals of horrible sanity" --ar 16:9 --v 6
Improved Prompt Accuracy
With the improved prompt accuracy in v6, you can be MUCH more specific about what you want in the image and it will listen to you (most of the time).
Prompt: /imagine top down wide angle view of a basket of blueberries, a basket of strawberries, and a basket of raspberries --v 6
For example, I ran this prompt below with v5.2 and v6 alpha:
/imagine two men walking down a city sidewalk. the man on the left has long black hair and is wearing a top hat, tweed jacket and red pants. the man on the right has blond hair and is wearing a blue tracksuit.
In v5.2, I sometimes got some of the details I asked for, but with v6 the prompt accuracy was much, much better.
Midjourney v6 Alpha Features
Midjourney v6 alpha does not include all of the same features that you are used to in v5.2, yet. Parameters such as --ar, --chaos, --weird, and --tile are there. Other features such as pan, zoom, and vary (region) will be available soon.
I am especially missing zoom out and style tuner at the moment, so I'm really looking forward to when those are available in v6. Vary (region) will also be really helpful for fixing poor text rendering.
Summary
Midjourney v6 alpha can draw text with some degree of consistency, has improved image quality (especially with realism), and has better prompt accuracy.
There are more features that will be released very soon. v6 will only get better.
Subscribe to my YouTube channel for Midjourney and AI tool tips and/or follow me on X/Twitter!