unsidhe

My first impressions of V 5 have not been good, I know that it has the whole world in love with its photorealistic quality, but I think that those of us who work with an illustrative or pictorial style have been disappointed compared to V4, which is still comparatively better in this sense.

I have a very similar feeling to when the test versions appeared at the time, which also had a more photo-realistic background compared to V3. I'm going to give it a chance because in theory this is not the definitive version but only the alpha.

From the beginning everything seems very nice in v5 when you start to test it, especially when you use simple prompts like character in the street that are the ones I usually use to test the style of summoning pictorial movements, aesthetics... and from the beginning spectacular results, the problem is when I ask for complex prompts.

And before you tell me that it's because I can't prompter in the new version, I have used v4 logic, it has followed to the letter the official indications of midjourney on this version, I have tried with sliders and the results don't convince me.

1.- Artistic materials lose texture in this version.

2.- Even if you are indicating illustration and material, it generates a photo because it wants to and either you mark those attributes with sliders or it ignores you.

3.- Many illustrations look like photos edited with some filters to look like drawings. They lose sharpness in the edges especially of the hair and you can't see well the stroke or the brushstroke.

4.- The moons look horrible, instead of being integrated with the rest of the style and they seem to be a source of luminosity.

5.- They look too much like the reference artist and when merging one of them it weighs more on the others which makes it difficult to mix references to generate an own style.

6.- The prompt with fantasy elements tends to ignore them in many cases and create a normal human being and a close element with the fantasy part instead of generating the fantasy elements.

7.- The image sets have less coherence between them which makes it difficult to get images that look like the same character without resorting to seed.

8.- Having a new photorealistic base, you can see that it has been trained by fashion photography, you can see it in the movement of the clothes of the characters, the tendency to generate too thin and too handsome characters.

In general I feel that I have less control when it comes to generating images, in v4 I give a series of concrete indications and I get both what is described in most cases and a mixture of all the styles, aesthetics and emotions that I incorporate to give the image the ambience. In v5 I tend to ignore things like the described clothes, gender, beards, artistic means... And following the indications of the own forums and material of consultation of the version.

And I get pretty pictures but they don't capture anywhere near the emotion, aesthetics or mood indicated.

So far it's been a frustrating experience and when you complain about it they're telling you that you're not describing it enough in the prompt. But I have tried with more extensive descriptions and with a natural language giving clear indications of what I want, I have tried to describe clearly character and scenario and then add words for atmosphere and style as in v4, I have tried with simple prompts and curiously is the one that gives better results, but is the one with which I have less creative control and I don't enjoy creating them.

In v4 the consistency of the images was higher so it was easier to put together a set of images.

I understand that it's an alpha version but at an artistic level the quality has declined, it doesn't read well the prompts introduced and offers less controlled results and of worse quality due to that lack of control.

The first time I tried Midjourney was around August 2022 and it was the beginning of something magical. At that time it was still a less developed tool than it is now, because it has only been a few months, but the leaps it has made in a few months are abysmal, it was in version 3 and it generated images that were not completely defined, with a very marked style and where you had to pray that it would draw the right number of arms.

And even with all these defects, I was amazed. Because for the first time my body was not a limitation to create art.

I was born with arthrogryposis, a disability that mainly affects my upper limbs, and although I have always liked to create, my arms have never allowed me to do it the way I wanted to. And I have always strived for it, but I was a soul demotivated by a bad art teacher, a frustrated photographer because I can't ask my arms to focus the angles the way I want them to?

That frustration was somewhat assuaged when I discovered graphic design and although on a functional level I'm pretty good at it, I was frustrated and ended up abandoning the more creative side because of how complicated and overwhelming it was to manage stock photos, and how complicated it was to get them.

So the first time I clicked on the discord it was something magical and gave me a result that fascinated me, I fell in love with it and fed the spirit of childish illusion that loved to sit and create, with the point that it also gave me very accurate images about what I wanted to express.

From that moment on it has been a relationship cultivated with hours of practice, creating palettes of styles, trying out words in simple concepts to see the style they bring and to get an idea of what it means to add it to a prompt, of trying crazy fusions and daring to experiment.

Because although the AI does the work of generating the art, it doesn't do it alone, it interprets your desires, your indications and knowing what to tell it so that it comes closer to what you have in mind is what requires a lot of work. For me it was somehow easy because having knowledge of CSS language was not difficult to understand its logic, but even so polishing my work to what I do today has taken me hours and hours of study and work.

I'm still half way through getting to know the tool, I'm still exploring many of the styling parameters and the photography or 3d rendering ones are still on the waiting list to be tested and create styling palettes with them.

But it's been an adventure to explore it all.