Exploring Midjourney v6.1 vs v6.0: What's Different?

Midjourney recently updated to the v6.1 model. This new version brings changes and improvements compared to the previous v6.0. If you're using Midjourney, it's helpful to know what these differences are. We'll look at image quality, new features, and how things like style references and character references work.

Image Quality Updates

With v6.1 being the default model now, many users notice an immediate change in image output. Here are some key points about the quality:

Overall Improvements

The v6.1 update generally boosts image quality. You should see fewer unwanted pixel issues and better-looking textures. Eyes and skin often look clearer and more realistic, especially in photo-style images. The model also does a better job with bodies and hands, making them more accurate.

The --Q 2 Mode

Midjourney v6.1 includes a --Q 2 option. The Q parameter affects quality. You can add --Q 2 to your prompt. Sometimes, this gives you more texture and detail. However, it might slightly reduce how well the image makes sense overall (coherence). This parameter existed in some older Midjourney models too.

New Upscalers

Version 6.1 comes with new 2x upscalers. Upscalers make your image larger. When comparing Midjourney's new upscalers to external tools, like Topaz Gigapixel, results can vary. Midjourney's upscalers can do a good job, sometimes fixing small details like eyelashes or hair texture. However, a tool like Topaz might handle skin texture, sharpness, and eye clarity better in some cases. Many users switch between Midjourney's upscaler and other tools depending on the specific image and what they want to do with it. Testing different upscalers helps you find the best result for your art.

Conventional Beauty Bias

Some users have noticed that v6.1 tends to create images of people with smoother skin and fewer perceived imperfections. This means there's a tilt towards what is often considered conventional beauty. Midjourney used user data to improve quality in 6.1, but this had the side effect of increasing this bias. Midjourney is aware of this and is working to address it in future versions like v7. If you want to create images of people who look more average or distinct, you might find v6.0 currently works better for those results.

Character and Style References

Using consistent characters and applying specific styles is important for many creators. Here’s how v6.1 handles these:

Character References (--cref)

You can use your existing character references with v6.1. You can also create new characters in v6.1 and use them in v6.0. Version 6.1 generally handles character references similarly to 6.0. You might see the conventional beauty bias appear slightly depending on the reference image you use, but the core functionality is the same. There hasn't been an update to the character reference technology itself.

Style References (--sref)

Existing style reference codes work in v6.1. Midjourney v6.0 and v6.1 share the same basic visual style understanding. However, you might see differences in the output. Images generated in v6.1 using an sref code might appear less detailed or less intricate than those generated in v6.0 with the same code. The style will be similar, but the level of detail can differ. You can still use your favorite sref codes, but be ready for slight variations.

Improved Text and Tile

Two technical improvements in v6.1 relate to text and patterns:

Text Rendering

Midjourney v6.1 is said to have better accuracy when adding text to images. It might handle longer or more complex phrases with fewer errors. If you use this feature, you can experiment to see how much of an improvement there is.

Tile Parameter (--tile)

The --tile parameter creates seamless patterns. In previous versions, upsizing a tile image often broke the seamless edges. This issue is fixed in v6.1. The tile parameter works better now, and the seamless pattern holds up when you use the new upscalers to create a higher-resolution image.

The New Model Personalization System

Version 6.1 introduces an updated model personalization system. This system tries to understand your preferred visual style based on how you rank images. It aims to be more precise in capturing your aesthetic.

Your personalization code, which represents your style preferences, changes as you rank more images. You can see all your current and past personalization codes using the /list personalized codes command in Discord. You can still use codes generated in v6.0 in both 6.0 and 6.1. New codes generated after the 6.1 update can also be used in 6.0. You can combine multiple personalization codes, but they need to be from the same system era (you can't mix old 6.0 system codes with new 6.1 system codes).

Testing your old and new personalization codes across both v6.0 and v6.1 models is a great way to see which combinations you like best. The differences might be subtle in some prompts or more obvious in others. Having multiple codes gives you more ways to experiment with your style.

Managing different model versions, settings, and personalization codes can become a lot to keep track of. A tool like the Midjourney Automation Suite from TitanXT can help simplify generating images with consistent styles and characters across different setups.

Getting the Most Out of Midjourney

Midjourney continually evolves, and understanding the updates helps you get the best results. While v6.1 brings improvements in quality and technical features like text and tile, some users might still prefer the creative feel and detail of v6.0. The conventional beauty bias in 6.1 is a known issue that is being addressed.

Experimenting with the different models and personalization codes is key to finding what works best for your specific creative goals. Don't be afraid to try out the new features and compare them directly.

Whether you stick with v6.0 for certain aesthetics or embrace the new features in v6.1, having efficient ways to manage your prompts, styles, and character references is essential. The Midjourney Automation Suite from TitanXT is designed to streamline your workflow, making it easier to manage complex tasks and generate images more efficiently.

Final Thoughts

Version 6.1 offers notable advancements, but v6.0 still holds value for its specific aesthetic and diversity in certain image types. The best model depends on what you are trying to create at the moment. Keep experimenting, and see how these different versions can serve your creative process.

Ready to take your Midjourney workflow to the next level? Check out the Midjourney Automation Suite and see how automation can enhance your creative output.