V5 is Live!..... and office hours 15/03/2023

Is it a bird? Is it a plane? No, it's Midjourney V5! Well, the alpha V5 anyway. Just a few hours ago, V5 was prematurely released and crashed within the first 20 minutes or so, but it was re-released about an hour later and has been going strong. I mean, have you seen the hands?!

These are initial generations zero rerolls. 

/imagine prompt: a handshake

/imagine prompt: a hand with the number 5 on the palm

To access V5 just type --v 5 at the end of your prompt or you can change versions in /settings.

 V5 has all aspect ratios and --tile, and our old friend Image Weights are back in rotation, making it much easier to strengthen image prompts.

Some noteworthy news and views

- Everything is already upscaled by default. So U1-4 is just separating the images.

-We were warned that prompts from older versions may not work as expected in V5. 

-V5 is also less opinionated, so we will be getting exactly what we ask for, for better or for worse. Words in V5 are super powerful but you also need to give instruction on what you want to see. We need to crank up that prompt craft, use rich expressive language that tells MJ what we want. We also have the --stylize parameter in our back pocket we can crank up in an emergency.

-In fact, the default for V5 is photographic so to get art styles you are going to want to put style names in the prompt then add --stylize 1000 to the end of your prompt. 

-Midjourney has also been working on it's language processing so writing prompts in sentences instead of lists in V5 may be the way to go.

 While V5 is still in the early stages and requires further development, we have access now, and it is precisely what they promised it to be: sharper, clearer, more coherent, more realistic, but different. Do we like different? Sometimes. Do we like this different? It has only been a few hours, so we will see.

Lets go through some before and afters

/imagine prompt: woman fairycore -- v 4

/imagine prompt: woman fairycore -- v 5

So you can see extremely photographic as default.

As I said it's going to give you what you ask for, so if you want Botticelli that is what you are going to get. 

/imagine prompt: Botticelli woman --v 5

This was after fleshing out the prompt a bit and pushing stylize all the way up

/imagine prompt:  beautiful Botticelli woman with long flowing hair, full color, digital art --stylize 1000  --v 5 

/imagine prompt: woman synthwave --v 4

/imagine prompt: woman synthwave --v5

And of course, we have --tile back. I know a lot of people were excited for the return of this feature.

/imagine prompt: elegant minimalist floral pattern with white background --tile --v 5

So everyone is going to have to put effort into their prompts that pretty filter has been turned way down.

Some of the old issues still haunt the algorithm

Terrible text

MJ still doesn't direction well when assigning specific colors.

We still have a bit of a blending problem

/imagine prompt: rabbit and tortious --v 5 

I am really looking forward to digging into V5 and seeing what it's capable of.

Fair warning, I waited a long time for these images to generate on relaxed, this is probably because of the default upscale.

 Oh and office hours was mostly about V5. David spoke about aliens and life in space. I am happy just going where no man has gone before in V5 thank you very much. 

They are working on a beta upscale for V5 and that /describe image to text feature is coming soon. 

I am off sown the rabbit hole, 

See you around wonderland.