If computers can imagine things, it's all thanks to AI. And with the advancements of computation, the technology just gets better and better.
Before OpenAI created ChatGPT, there was the time when people were awed by how AIs can create photos of people that don't exist, and how AIs can generate "perfect" human faces, and many more.
There is even an AI purposefully designed to create nudes that don't exist.
Then, things escalated when OpenAI created DALL·E and DALL·E 2.
While AIs got better, they tend to have quite a few issues. And the traits are somehow identical from one AI to another.
Midjourney solved some of them, brilliantly.
Starting today our community can test Midjourney V5. It has much higher image quality, more diverse outputs, wider stylistic range, support for seamless textures, wider aspect ratios, better image prompting, wider dynamic range and more. Let’s explore!
— Midjourney (@midjourney) March 15, 2023
Through Midjourney v5, the popular AI image generator's AI can produce strikingly lifelike visuals. To make this happen, the v5 model uses cutting-edge tools and a updated neural architecture.
This allows it to generate astonishingly photorealistic aesthetics and designs.
In what took five months to complete, the v5 model can create proper human hands, for example.
This is possible because Midjourney v5's aesthetic range is significantly greater, allowing it to understand user requests more accurately. The quality is also improved, with Midjourney saying that the v5 has twice as much resolution and a wider dynamic range that its predecessor.
It's more detailed too.
Because of this, the AI should be able to generate five human fingers per hand, and doesn't add more teeth than needed.
The developers also claimed to have "dramatically" enhanced facial detailing.
As a result of this, v5 allows a much higher level of graphic detail than any other image generators can ever make at the time of its launch.
The team announced the news on the project's official Discord channel.
Other improvements that come with v5, include:
- Much wider stylistic range and more responsive prompting.
- Much higher image quality, with two-times resolution increase, and improved dynamic range.
- More detailed images, with details that are more likely to be correct, and with less unwanted text.
- Improved performance with image prompting.
- Supports for seamless tiling (experimental).
- Supports for aspect ratios greater than 2:1 (experimental).
- Supports for weighing image prompts versus text prompts.
Initially introduced in beta, Midjourney's v5 is still under heavy development.
The creators cautioned that two-word requests users make on v4 may not be as effective in v5, and that users should instead query longer, more specific prompts. And because it's still in betathe developers also cautioned that the neural network may undergo significant changes before the final release.
It's worth noting that because v5 is bigger, it's also slower.
By default, it works with resolution of 1024x1024, meaning that the AI art generator may take a bit longer to churn out its creations than its predecessor. However, upscaling is immediate.