The new Midjourney 5 update takes photorealism to a whole ‘nother level

Midjourney 5 Unveiled
Honor X9b Ad
Honor X9b Ad

These days, almost everyone has heard of ChatGPT (Chat Generative Pre-training Transformer). And its latest iteration, GPT-4, can now respond to both text and image queries. Although its response is still text-based. But what if you wanted to transform your queries into images? That’s where AI art tools like DALL-E 2, Stable Diffusion, and Midjourney come in. And the latest Midjourney 5 (v5) update fixes some of the most pressing issues with such generative tools, to deliver incredibly photorealistic results.

What is Midjourney?

First, let’s learn a little more about Midjourney. It is an artificial intelligence (AI) art tool (or an image synthesizer) that is trained on an AI model based on millions of existing human-made artworks. And based on those models, it responds to user prompts with images. David Holz, a former student researcher at Max Planck, is the founder of Midjourney. Unlike most other AI tools, Midjourney responds to prompts via bot commands in Discord. Its main rivals are DALL-E 2 and Stable Diffusion.

How is it different from DALL-E 2?

Midjourney generates artistic and visually appealing photos, whereas DALL-E 2 tries to provide us with actual photos most of the time. Check the images below for a better idea.

What’s new in Midjourney 5?

The new Midjourney 5 update—which is currently only available to paid subscribers—takes its AI capability to the next level. With Midjourney v5, you will get even more realistic and detailed images than with version 4. Midjourney has previously faced criticism for the way it handled human hands and teeth, as it depicted 7 fingers in one hand and a garishly unnatural number of teeth.

However, these issues have been addressed in version 5.

Midjourney 4 vs 5
Source: Ars Technica

For the best results, Midjourney v5 requires you to type in a more detailed prompt in a proper sentence rather than just a bunch of words separated by commas. Version 5 also features improved blending. Midjourney used to simply join two images when asked to blend them, but now the blending is nearly perfect.

But texts in images still remain a challenge. If you want to add text to your image accurately, you must use Photoshop or other tools. According to Midjourney officials, version 6 will be released within the next two months.