How to use image weights in Midjourney
My clay Minime

How to use image weights in Midjourney

Consistency and replicability in Midjourney is hard and takes a lot of practice. One way to achieve it is by using image prompts, which means not only do you input a text prompt but you also provide one or more images to the AI model.

There's a very handy parameter called "image weight" in Midjourney that allows you to control how important, or prominent, the provided image should be in the end result. To give you an overview of how this works, I first uploaded a profile photo of myself (which was generated by the AI Lensa, but that's another story). Here it is:

No alt text provided for this image

I then used this prompt to get us started:

/imagine <URL to the above image> clay stop motion figure        

As you can see from the image below this gave a wide spread of results where some variants look a lot more like clay figurines than other. And since I was going for a specific style, I need to do something about this. I could try refining my prompt, using additional strong words, but fortunately there's also another solution.

No alt text provided for this image
A wide spread of results, some more clay-like than others

The parameter image weight (--iw) accepts a value between 0-2, where 1 is default. To showcase the result of different values, I use the very handy permutation prompt feature that looks like this:

/imagine <URL to the profile photo> clay stop motion figure --iw { 0.5, 0.75, 1, 1.25, 1.5, 1.75 }        

This will generate six prompts with incremental image weights. Here we go.

Image weight 0.5

No alt text provided for this image

Image weight 0.75

No alt text provided for this image

Image weight 1.0

No alt text provided for this image

Image weight 1.25

No alt text provided for this image

Image weight 1.5

No alt text provided for this image

Image weight 1.75

No alt text provided for this image

The impact of this quite powerful parameter is clear, and it serves as a very useful way of controlling the style of your output, or for replicating the likeness of a photo of a specified person.

That concludes our tutorial for the day, but you're getting some more images from me toying around with my profile photo even though you never asked. ??

No alt text provided for this image
No alt text provided for this image
No alt text provided for this image
No alt text provided for this image

Other Midjourney articles I have posted

Are you keen on trying out Midjourney yourself? Here are some other articles I've published on LinkedIn for additional tips and tricks.

Drausio T.

Award-Winning Creative Leader | Social +Technology driven creative with 25+ years delivering results (and awards) to global brands | Ex-WPP, FOX, MTV

1 年

Which profile of Lensa did you use?

回复
Lukasz Lindell

Senior Policy Communications Lead TikTok Nordics

1 年

Love this, thanks!

Cameran Harman

Strategic Partner Manager, Global Sports Partnerships at Meta

1 年

are there action figures available? I feel like I need the one weighted 0.75 on my desk.

Nick Wrenn

News leader and team builder, using digital innovation and partnerships to grow audiences. Ex-CNBC, Meta, CNN, BBC in UK and US.

1 年

Thanks for the tutorial John. Superman is the best.

Sebastian Tibbling

?? Byr?pappa p? Roy, a_delivery, Oh Charlie och Adviser @ Another Media Group

1 年

Snygg du ?r i lera! ??

要查看或添加评论,请登录

John Severinson的更多文章

社区洞察

其他会员也浏览了