larger images, video games, video generation, fingers, what we expect from the next version of AI
Several leaks have occurred about Midjourney v6, a new version of the automatic image generation artificial intelligence. After an impressive v5, the changes coming soon could also be important. Overview of what could change on Midjourney.
Last March, Midjourney v5 was released, a new version of an artificial intelligence image generation model. It brought a precision never seen before on images created by a computer, with richer and more detailed renderings. In a while, Midjourney v6, its successor, could be released with notable improvements that could set it apart from competitors like Stable Diffusion or Dall-E.
When will Midjourney v6 arrive?
At the moment, we do not know much about the schedule of Midjourney, the independent research laboratory created by David Holz. The v6 of the tool has not even been confirmed yet. But journalist Eray Eliaçık from dataconomy predicts a release in the next two months. However, he warns that this date is likely to change depending on the progress of development and testing before public availability.
[see_More slots= »1682025″]
New features that could arrive on Midjourney v6
Several features and improvements are expected on the future version of Midjourney, here they are.
Better defined images
Eray Eliaçık claims that Midjourney v6 will generate images with a maximum size of 2048 by 2048 pixels, where today users are limited to 1024 by 1024 pixels. Which could give images four times larger than currently.
The advantage is that they can be used more easily: social networks, wallpapers, etc. Indeed, most screens today have higher definitions than what Midjourney offers. It remains to be seen whether it will actually be possible to have such images: this would require significantly greater computing power. But this is one of the most anticipated improvements.
A better understanding of requests
This v6 could also understand more precisely what it is asked to generate. What allow more realistic and more adequate images. It could also be a way to democratize the tool: at present, you have to use a certain punctuation, place your words in a certain order. All of these hurdles cause many to either not try Midjourney or fail to achieve what they want. Midjourney v5 had already made it possible to use a more natural language: this could go even further in the next version.
Choose the level of image variation
Other rumors indicate that v6 will offer greater customization of image variants. When you make a request to Midjourney, it makes four proposals. Soon, we could precisely choose the level of variety that we want in the proposals, so that they more or less approach the request.
Also, the changes you want to make on an image could be more precise, by changing a single detail for example. Enough to keep an image, by modifying only a color, a texture, an object, the sharpness of an element, etc.
Generating 3D models with Midjourney v6: what appeal to video games?
Another very important feature that could land: the ability to generate 3D models in several formats. What help some users, including virtual reality professionals, 3D printing enthusiasts, video game developers, etc.
Other 3D modeling generators exist, such as OpenAI’s Shap-E, which is still a prototype. Using a simple textual query, the model manages to create 3D shapes. At the moment it takes hours, but over time the generation could be faster and faster.
The end of hands with six fingers?
One of the big limitations of Midjourney v6 is the creation of hands, and more specifically fingers. Between the phalanges, the shapes of the fingers, their position, everything is there to deceive the AI.
This is also one of the most effective ways to know if an image has been artificially generated or not. This is the problem with all image-generating AI models and Midjourney is no exception. It is clear that efforts had been made with the arrival of Midjourney v5, but v6 would go even further. The relayed leaks indicate that the improvement in the hands would be significant.
What if Midjourney v6 started generating videos?
Midjourney v6 might also start generating videos. It would not be the only model that would be able to do this. Meta has Make-A-Video, which can create videos using text description.
As with 3D models, it’s still very limited, but within a few years it could be much more widespread.
When is an API for Midjourney?
The image generation tool has a big limit for the moment: its use. To create images, you must go through a Discord bot to which you make requests. No site, no app. One of the most repeated requests from users is to be able to use Midjourney wherever they want. An understandable request: this could increase the uses of Midjourney tenfold.
At the house of CssTricks, we could use the tool directly from the article writing page. It could also be integrated into office suites: Microsoft 365, Google Docs, Canva, etc. What also goes to compete with Adobe Firefly, Adobe’s generative AI, capable of enlarging images.
To do this, you would have to use one thing: an API. As a reminder, this is a piece of code that allows you to connect two software, sites, applications. From one, he calls on the services of the other. In our case, the destination service needs an API to ask Midjourney to generate an image, and it is necessary to receive the proposals.
one day
— Midjourney (@midjourney) May 6, 2023
For the moment, the arrival of the API has not been confirmed. But on Twitter, the laboratory account explained last May that this would be the case “ A day “, without further details. As for a web interface for Midjourney, the lab said not knowing when it would arrivewithout indicating that it would exist one day or not.
The Watt Else newsletter is THE unmissable event dedicated to the mobility of the future. Register here!
