Midjourney V6.1 Is Finally Here

The long wait is finally over. After months of anticipation, Midjourney announced the release of V6.1.

Seven months since the first official release of the Midjourney V6 model, the most popular AI image generator is finally back with some much awaited upgrades. The creators admit that this release has been in the works for a long time and has been guided by the priorities submitted and ranked by the community.

Contrary to usual versioning practices, Midjourney did not release the new model as V6.5. Instead, they opted to roll it out in two phases: V6.1 and V6.2.

Why?

According to David Holz, the creator of Midjourney, they need to gather as much data as possible on people using V6.1 before they release a V6.2 model in the next month or so.

What’s New in Midjourney V6.1?

Here’s a complete list of changes and improvements in V6.1:

• More Coherent Images: Improvements in rendering arms, legs, hands, bodies, plants, and animals.
• Enhanced Image Quality: Reduced pixel artifacts and better textures, including skin and 8-bit retro styles.
• Detailed Small Features: More precise and accurate small image features like eyes, small faces, and distant hands.
• New Upscalers: Improved image and texture quality with new upscalers.
• Speed Improvements: Roughly 25% faster for standard image jobs.
• Improved Text Accuracy: Better accuracy when drawing words via “quotations” in prompts.
• New Personalization Model: Enhanced nuance, surprise, and accuracy in personalized images.
• Personalization Code Versioning: Ability to use any personalization code from old jobs to utilize the personalization model and data from that job.
• New — q 2 Mode: A mode that takes 25% longer but sometimes adds more texture at the cost of reduced image coherence.
• General Aesthetic Improvements: Overall, images should look “generally more beautiful.”

Example Images of V6.1

I wont be doing an extensive test of the new model on this article. Instead, I will do some examples to see how these improvements play out, especially focusing on hands and text rendering.

Rendering Hands

Prompt: an image of crowd of people raising their hands — ar 16:9 — s 750 — v 6.1

Okay, most of the hands and limbs from this photo are looking right but a few still come out a bit off. I’ll switch back to V6 and generate an image with the same prompt.

It’s worse than the previous image. Some hands are flipped, and others don’t even have complete fingers.

Rendering Text

Prompt: A white paper card on a wooden table with a text “Midjourney V6.1 is awesome” written on it — ar 16:9 — s 750 — v 6.1

Despite V6.1’s promise of improved text accuracy, it occasionally excluded parts of the text, such as “V6.1.” Surprisingly, V6 performed better with this prompt, rendering the text correctly.

Am I missing something?

Close-Up Shots

Prompt: 14h A close-up image of a person with their eyes closed and hands on her face, a serene expression on their face. Their head is slightly tilted back and encircled by a soft, glowing blue light, symbolizing the calming theta waves. The background is a soothing gradient of deep blue and purple hues, resembling a tranquil night sky. — ar 16:9 — s 750 — v 6.1

I can’t really say for sure if there’s any improvement because V6.0 has already proven itself to be really good at rendering small and finer details.

Here are more examples shared by David Holz from Midjourney.

The new --q 2 parameter refines details and enhances the image’s overall quality, but it’s not always guaranteed. Check out this example from X user Dogan Ural.

How to Enable V6.1

To enable V6.1, access the Midjourney bot and bring up the settings page by typing /settings. In the model dropdown, select V6.1.

You can leave the rest of the settings on default.

For users on the Midjourney website, switch to the new model by navigating to the settings page and setting the model version to 6.1.

Note that V6.1 currently does not support new inpainting or outpainting models. If you use features like zoom, reframe, repaint, or vary regions, the system will revert to the V6.0 model.

Future Plans

Here are some insights into Midjourney’s upcoming plans:

We will likely release a V6.2 model in the next month or so with further improvements but first we need to gather as much data as possible of people using V6.1. So as of today we’re not just releasing V6.1 but we’re also making it the default model on Midjourney for all users.

What we can expect in V6.2:

• Improvement: Better rendering of bodies, hands, and text.
• Image Quality: Significant enhancement in overall image quality.
• Model Aesthetic: They are considering making the default model aesthetic more gender-balanced, possibly in V6.2.

If you want to share your feedback, go to Midjourney’s Discord channels and share your thoughts or report bugs.

Okay, that’s about it. What do you think of V6.1? In the next article, I will compare V6 and V6.1 in greater detail to see how much of an upgrade users can expect with this new model.

Final Thoughts

Overall, V6.1 is a welcome upgrade, though the differences from the already impressive V6 model are not groundbreaking. In fact, in some of the examples I showed above, the older model performs better than the new one.

Hopes are high for V6.2 to address current limitations, especially in human anatomy and text rendering. I can’t wait to see more exciting new releases and upgrades in the coming months. Video or 3D next, please?

And a notable mention: Elon Musk recently followed Midjourney on Twitter. Could this mean Midjourney on Grok? Only time will tell, but it’s an exciting prospect for the MJ community.

‍