Welcome to this week’s edition of Architecture Insights.

With yesterday’s release of Midjourney v6 comes an opportunity to review its capabilities for architects and designers while taking a look at how far the AI text-to-image generators have come.

This Past Week in AI

1. Midjourney releases Version 6.0.

Summary: We will get into this in the main section. You can also read the full overview from Midjourney.

2. OpenAI releases a full guide on prompt engineering. (Read here)

Summary: This is an extensive guide that applies to more than just architects but to everyone who uses ChatGPT. 6 different strategies are outlined to help with achieving better results.

Midjourney Version 6

It has been over a year since the release of Midjourney into the world of AI image generators and architects have been testing and iterating with every version since.

Midjourney version 6 brings about a good opportunity to 1. review its capabilties, 2. compare it to other AI text-to-image generators and 3. explore how far AI has come in enhancing the workflows and possibilities architects have with AI.

Here are some key things to note with the recent update of v6:

• Improved model knowledge

• Improved upscaling

• V6 is slower and more expensive as a result of its upgrades

• Using text in your prompt such as “photorealistic, 8k, 4k” is useless.

Here is how to start using V6 on discord (It is very simple):

Type in /settings, enter, and select v6 from the options, that is it.

Alternatively, just add --v6 to the end of your prompt.

Let’s try out using the same prompt with v5.2 (the current version of midjourney) and v6.

Prompt: A 20th-century modern-style gas station with curved lines, a flat roof, and panoramic glass windows.

Version 5.2:

Version 6.0:

These are very interesting results.

On the one hand, v5.2 looks more pleasing and features more details with its lighting and overall design.

v6 just looks slightly more realistic in its design and image resolution.

This raises the question: how does using midjourney v5.2 compare to v6.0 for architectural designers?

When it comes to choosing between two concept renderings that we might show to a client for feedback, many of us might prefer the v5.2 image because we think it is more visually appealing than the v6.0 image.

It's unclear where Midjourney is going for designers with the recent update, but the generated images appear more "realistic".

Perhaps with continued experimentation with v6, Midjourney may guide us towards creating more detailed prompts that wouldn't have worked well with previous versions.

v6 is only a day old at the time of this post.

Version 5.2 prompting system lacks clarity and leaves much to the imagination of midjourney for “filling in the gaps”. Both Version 5.2 and the new Version 6.0 provided the requested results, but Version 5.2 offered additional information. We hope to fully understand the capabilities of Version 6.0 once we become more familiar with its prompts.

The buzz around the internet keeps saying that prompting will need to be more detailed and exclude any unnecessary “fluff” if you want the best result from v6.

This update has the potential to lead us closer to a world where renderings can be fully created and customized inside midjourney, giving us the same capabilities as photoshop or other rendering software.

For now, we can continue to experiment with different prompts to test out where v6 may be better and where it may still be lacking for its use in creating architectural renderings and images.

Resources & Tools

The new OpenAI ChatGPT guide:

AI Image of the week

Thank you for reading this week’s issue, check past issues here. Share this newsletter with colleagues, friends, or anyone interested in the combined world of architecture and artificial intelligence.

Feedback is always appreciated, let us know how we did by responding to this email.

Until next Friday,

A.I.