Text: Shane
Edited by Tang Yongyi
After 9 months of waiting, finally, the latest version of Midjourney, the king of Wensheng map, is launched today in V6 raid!
How sudden was the launch this time? so sudden that I haven't finished training yet...... Midjourney V6 is still in the emergency RLHF, and the official said that the second community evaluation has been opened. Users need to make annotations, annotate 200 sheets, and only send 1 hour of use time.
图源：Midjourney
Since the release of Midjourney V5 in March 2023, netizens have been waiting for 9 months, even if they have to mark it to play, it will not affect the enthusiasm of netizens-
Now, you can use the /setting command to select the V6 Beta version in the model selection box.
图源：Midjourney
The first netizens to try the new have produced the first wave of pictures:
For example, in this picture, Musk is holding a goblet and looking at it carefully, can you tell if it was drawn by AI?
Source: Twitter
Another example is this, the Pope in the coat of the Parisian world.
Source: Twitter
To talk about the effect of this Midjourney V6 update, it is simply as follows:
The distinction between an image and a photograph has disappeared.
Seeing such a miraculous result, many netizens exclaimed: "It's amazing!"
Source: Twitter
Source: Twitter
Accommodating more tokens, stronger semantic understanding
According to the official statement, the current V6 Beta update includes:
More accurate cue following and longer cues, improved coherence and model knowledge, improved image cues and remixing, smaller text drawing capabilities, and improved upscaler with 2x higher resolution.
The simple summary is more words, better semantic comprehension and image generation, more personalized design of images, slight text-drawing ability, and so on.
In terms of image quality, Midjourney V6 has improved by leaps and bounds compared to the past. The following is a comparison of V5 and V6 pictures, and it can be found that the pictures generated by V6 are very good in terms of material texture, light and shadow performance, and structural authenticity.
In addition, the expression of composition and color is also a step further.
Source: Twitter
Source: Twitter
In addition to the great improvement in image quality, Mimjourney V6 also has a more intuitive understanding of text prompts.
According to designer Tatiana Tsiguleva, Midjourney V6 can understand the user's prompts more accurately, which means that there is no need to type in those strange phrases or words, such as tag, photorealistic, 4k, etc. Here's a simple structure based on a typical language framework:
图源：Midjourney
Now, with V6 you can specify any detail, or use language to control the composition, or even add text to the image.
Source: Twitter
From hairstyles, beards, rackets to the folds of clothes, the details are all on the line.
Source: Twitter
The hairy roots of cats are clear.
Source: Twitter
Didn't see enough?
Source: Twitter
Source: Twitter
Source: Twitter
Source: Twitter
It was launched before Christmas, and the official website was in urgent trouble
Sure enough, after reading the first wave of pictures hand-rubbed by netizens, Midjourney is still firmly on the throne in the field of AI maps.
Let's just say, when can you use pinching?
No, the good news is coming, some netizens said that Midjourney V6 is in the final round of RLHF, and after the style fine-tuning is completed, it will be released before Christmas.
According to official information, V6 is likely to be launched today and tomorrow.
Source: Twitter
As soon as the news came out, netizens brushed and said: I am the king of urgency, I can't wait!!
Source: Twitter
Source: Twitter
Some netizens bluntly said:
圣诞节礼物，就看你了，Midjourney。
Source: Twitter
Finally, some netizens broke the news that Midjourney V6 seems to be able to generate even paper pictures... In any case, this wave of V6 improvement is really praiseworthy, just wait for the time to start playing (flies rubbing hands).
Source: Twitter
Source: Intelligence emerges