[v5-related] Midjourney v5 and v4 generation effect comparison

Midjourney v5 is finally here, and this article will share with you the latest updates and enhancements in this version compared to the previous version v4. Fasten your seat belts and explore the new changes of Midjourney v5 with us!

More realistic output images

In addition to producing more realistic and detailed images, the new models require more precise instructions. They employed new neural architectures and new aesthetic techniques to develop the model, which took about five months to train on Google Cloud's "AI supercluster," Midjourney said. This model was able to produce more realistic images with more accurate details. In the first test, with the same prompt without an explicit image style, v4 produced an illustration, while v5 produced a more realistic museum installation for the prompt "tree made of banknotes". The result is also closer to the expected result thanks to v5. Midjourney v4:

Midjourney v5:

Sharper, more detailed images

Since the Midjourney upgrade, you'll notice that your images are now sharper and more detailed. This is due to the benefits of new image processing algorithms implemented in this latest release. As you can see when comparing the V4 and V5, the V5 produces a sharper, more detailed image. Here is a comparison of V4 and V5 for the prompt "Two Toddlers Playing with Legos". At first glance, the images of the V4 look real. Midjourney v4: Midjourney v5: On closer inspection, however, the images look more like cartoon characters, with children's faces looking like rubber masks, their hands a mess, and Lego bricks looking like blurry replicas. The V5, on the other hand, is more like a real photo, and the level of detail is truly amazing.

Wider range of supported styles

The Midjourney team says the V5 has five times the style range of its predecessor, which means you can use longer, more descriptive prompts and write full sentences to inspire a wide variety of visuals, from landscapes to architecture and more . The algorithm is also optimized for natural language processing, resulting in more accurate results. In addition, many issues with V4 have been fixed, such as better handling of large numbers of people, more realistic hands, more accurate number of fingers, etc., while also reducing image artifacts.

Use image cues to generate more "accurate" and predictable results

Midjourney V5 offers a wider variety of images than its predecessors and is no longer limited to a specific art style. Holz suggests that to achieve the desired effect, longer cues need to be used to clearly describe details in the image, such as lighting and atmosphere. Holz said the current V5 model is an alpha version that will undergo significant changes before its final release. Like the previous version, the final version will have a more beginner-friendly default style. Creatives should not expect to see the current v5 style in the future. Biomorphic skyscrapers and snowy mountains:

MidJourney V5 higher resolution

Users of MidJourney have been asking for an upscaling of image resolution, and we're happy to announce that the team is gradually making this possible. By default, MidJourney V5 images have a resolution of 1024 x 1024 pixels, compared to 512 x 512 pixels for V4. V4 images had to be upscaled to reach this resolution, so this is a significant improvement. In the future, we may see an upscaler for the V5 that can increase the resolution to 2048 x 2048 pixels. However, now, 1024 x 1024 pixels is the default and maximum size.