Shulou(Shulou.com)11/24 Report--

The original title: "lifelike to tremble!" Midjourney draws a perfect Chinese couple, painters, actors and models are eliminated with one click. "

After GPT-4, Midjourney V5 is online. Netizens tried it one after another, drawing a pair of Chinese lovers, visually exploding and crushing human painters.

Yesterday, a Chinese couple painted by Midjourney V5 caused an uproar on the Internet.

Many people exclaimed, "Human painter, stop living."

Some people do not believe that this is the work of Midjourney, but also experienced it. After entering the following prompt, the painting style was instantly amazing.

"A pair of young Chinese lovers, wearing jackets and jeans, sitting on the roof, the background is Beijing in the the background is Beijing in the, and the opposite building can be seen-- v 5-- s 250-- Q 2"

After the release of GPT-4, Midjourney also updated V5, which not only makes the drawing more realistic, but also makes prompt more sensitive.

Most importantly, the long-criticized "AI can not draw fingers", Midjourney V5 has been pinched.

There is also a full screen of movie texture, in addition to OMG, or OMG.

Some netizens even exclaimed: models no longer exist!

With the effect exploding, a wave of designers are about to lose their jobs, from Midjourney's V1 to V5, you can see how fast AIGC has evolved over the past year.

The latest V5 version has higher image quality, more diversified output, wider style, seamless texture support, wider aspect ratio, better image prompts, and wider dynamic range.

Excited heart, trembling hands, domestic netizens have started to exercise a wave.

The first photo photo to start with:

Although V5 can draw 5 fingers, it is almost meaningful in the details of the fingers.

For example, try to get Midjourney V5 to take a picture of Van Gogh's room:

When Midjourney V5 paints the appearance of Chinese beauties, the control is also very in place.

V5 can also grasp the beauty of Chinese classical style, draw a graceful beauty, full of details and granularity.

Some netizens created the movie "the Rebirth of Hangzhou" with GPT-4+Midjourney V5, and even the introduction page of Douban appeared, which seemed to be the case.

In December 2022, Carpenter Tan held a comb competition. Human designers designed combs as follows:

Netizen Huang Zhongsheng tried a wave with V5 and was surprised to find that Muggle could not play at all in front of AI.

When others reported that the combs above were so beautiful that they could only be used for collection, netizens made a whole wave of daily combs.

And combs for children.

In foreign countries, some netizens directly used Midjourney V5 to draw the texture of the The Matrix film, which made people wonder if it was a screenshot in the movie.

And presented prompt:

"Keanu Reeves in The Matrix. AI cinema. Synthography. Cinematic. Film grain. -- ar 16:9-- v 5 "

Some netizens tried to use the classic meme of "smiling Girl at the scene of the Fire" to ask Midjourney to create "Loki at the scene of the Castle Fire".

Let's take a look at the effect of this explosion.

Netizens also used their brains to add Loki elements to the picture of "Beijing Roof couple".

There is also a funny version, the tornado is coming, and the old lady is riding an electric donkey.

The facial features of Asians are also very realistic.

Travel bloggers don't have to go out, just train Midjourney at home.

Some netizens made a comparison between V4 and V5. By entering a picture and a text hint, the picture generated by V4 cannot be compared with V5 in the shadow treatment of objects in light.

Here's another close-up of the eyes:

The output pictures of the three versions of V3, V4 and V5 are compared by entering the same prompts and settings.

In the V5 version, the sparkling surface of the water, as well as the reflection of fish swimming in the water, has reached the extent that it is fake and real.

But if you don't provide a hint of text about the environment, it generates a more real environment by default than V4 and V5.

Netizens can't wait to ask V5 to reproduce the movie style of Wong Kar-wai's Blueberry Night and find that V5 is more realistic in dealing with the reflection of glass.

And all kinds of works that are difficult to distinguish between true and false:

Musk, full of villain temperament:

Cat wears VR:

Where is it better than V4? With the latest upgrade of V5, one researcher made a table of the comparison between them compared to the previous version:

1. A secret shop (--tile argument) is used when creating a pattern. Previously, V3 had this parameter, but it was temporarily suspended in V4 and the V5 version was back online.

For example, enter "/ imagine tesseract pattern, repeat, white and blue highlights-- s 1000-- v 5-- tile" to generate the following figure:

two。 Aspect ratio (--ar argument) in V4, you can only have a custom aspect ratio that is not higher than 1:2 or wider than 2:1. In V5, you can customize the ratio to any width: length ratio, as shown below:

What does this upgrade mean?

The widest aspect ratio of 2.39 in modern movies can be achieved, you can make a movie screen, everyone can become a director!

V5 generates 2.4 VR 1 movie screen 3. Image weights can be customized (--iw argument) weights in image prompts can now be entered custom values. You can determine the extent to which the input image affects the resulting image.

Below, in Midjourney V5, enter a picture and a text prompt.

By customizing different weights, 2, 1, 0.5 generate different images.

The prompt entered is as follows: "/ imagine http://imgur.com/ romman.png a roman soldier with daisies in the background-- seed 123-- iw 2-- v 5"

4. Less opinionated "less opinionated" means that in V4, Midjourney outputs more creative images with fewer text prompts.

But this is not the case with V5, you may need to enter more adjectives and give more details, such as style or emotion, what the output image should look like. If there is no more description, Midjourney will output a default style, such as "photography" style.

This upgrade allows people to control the output of Midjourney more accurately through text prompts. It also means that there is "more cohesion" between input text and output images.

The above four upgrades are mentioned in the official blog, but the most unexpected thing is that Midjourney V5 can draw "5 fingers and toes"!

In addition, V5 has fine-tuned more realistic images / photos, so you can see more realistic images reflected from the output window and through the glass.

Take an examination of eyesight

Finally, here is a test question:

Of the following four pictures, only one is a real person, and the others are all drawn by Midjourney V5. Guess which one is the real person?

Reference:

Https://bootcamp.uxdesign.cc/midjourney-v5-is-here-so-whats-new-df3f3a645295

Https://twitter.com/mranti/status/1637326003927396352?s=46&t=iBppoR0Tk6jtBDcof0HHgg

This article comes from the official account of Wechat: Xin Zhiyuan (ID:AI_era)

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.