Midjourney is one of the most popular AI image generators. It runs on Discord and allows users to type in text prompts which in turn lets the AI generate an image that matches the text description. Midjourney has been constantly improving its AI models since its launch in March 2022. The latest version, Midjourney v5, was released in March 2023 and it brings some major improvements over the previous version, Midjourney v4.
In this post, we will go over some of the new features and benefits of Midjourney v5 and how it compares to Midjourney v4. We will also show you some examples of images generated by both versions using the same prompts.
Higher Native Resolution
One of the most noticeable changes in Midjourney v5 is that it produces images at a higher native resolution than Midjourney v4. The default output size for Midjourney v5 is 1024 x 1024 pixels, which is twice as large as the default output size for Midjourney v4 (512 x 512 pixels).
This means that Midjourney v5 images are sharper and more detailed than Midjourney v4 images. You can see more textures, colors, shadows, and nuances in the images generated by Midjourney v5.
For example, here is an image generated by both versions using the prompt: “portrait of a smiling young woman wearing biblical clothing in Jerusalem, brown eyes, photo realistic 8k”.
As you can see, the image generated by Midjourney v5 looks much more realistic and natural than the image generated by Midjourney v4. The facial details, skin complexion and clothes are all more defined and accurate in the the v5 image.
Unlimited Aspect Ratios
Another improvement in Midjourney v5 is that it supports unlimited aspect ratios for the output images. This means that you can specify any width and height for your image with the –AR command and Midjourney will adjust accordingly.
For comparison, Midjourney v4 only allowed square aspect ratio (1:1) at launch, which was later increased to 2:3 / 3:2 until finally supporting radios of up to 1:2 / 2:1.
This gives you more flexibility and creativity when using Midjourney v5. You can create images that fit any screen size or format without compromising on quality or cropping!
More Coherent Output
Another improvement that comes with Midjourney v5 is that it produces more coherent output than previous versions. This means that the images generated by Midjourney v5 are more consistent with the text prompts and less likely to have errors or glitches.
Some of the common issues that plagued Midjourney v4 were:
- Missing or extra body parts (e.g. fingers)
- Distorted or unnatural facial features (e.g. eyes)
- Inaccurate or mismatched colors (e.g. skin tone)
- Inconsistent or illogical lighting (e.g. shadows)
- Unrelated or irrelevant objects (e.g. logos)
Fortunately, these issues have been largely fixed or reduced in Midjourney v5 thanks to its improved AI model. The model has been trained on more data and refined with better techniques to produce more realistic and coherent images.
Here is an image produced in both v4 and v5, using the prompt “hyper realistic photograph, portrait of an African woman with a baby, colorful clothes, summer :: film grain, Kodak portra 800, f1.8, golden hour”.
More Realistic Human Hands
One of the most impressive improvements in Midjourney v5 is that it can generate more realistic human hands than Midjourney v4. Human hands are notoriously difficult to draw or model, even for professional artists. They have so many different complex shapes, movements, and expressions that can be hard to capture.
Midjourney v4 often struggled with generating human hands that looked natural and accurate. Sometimes they had too many or too few fingers, sometimes they were twisted or bent in unnatural ways, and sometimes they were out of scale or perspective.
Midjourney V5 has fixed most of these issues and can generate human hands that look almost indistinguishable from real ones. The fingers are correctly positioned and numbered, the joints are smooth and flexible, the nails are shiny and clean, and the skin tone is consistent and realistic. Just take a look!
Here is an image by the Midjourney team showcasing close-up images of human hands.
As you can see, the new model really nails it when it comes to creating the complexities and finer details on human hands. We’re not sure which prompt they used for these, but most images with hands in them will naturally be more coherent in v5.
How to Use Midjourney v5
If you want to try out Midjourney v5 for yourself, you will need to have a paid account on Midjourney’s website or Discord server. The current price for accessing Midjourney v5 is $9.99 per month or $99 per year. Sadly, v5 is not yet available to free trial members.
Once you have your account set up, you can use Midjourney v5 by adding “–v 5” to the end of your text prompts.
To use Midjourney v5 effectively, here are some tips and tricks:
- Use longer and more descriptive prompts to get better results. For example, instead of “a cat”, try “a fluffy orange cat sitting on a windowsill”.
- Use hashtags to specify certain attributes or styles for your images. For example, portrait #realism #fantasy #anime #watercolor etc.
- Use brackets to add captions or labels to your images. For example, [a man wearing glasses] {This is John}.
- Use parentheses to add comments or instructions to your prompts. For example, (a photo of me as an astronaut) (make me look happy).
- Experiment with different prompts and keywords to see what works best for you. Have fun and be creative!
Midjourney v5 is a remarkable upgrade over Midjourney v4 in terms of resolution, aspect ratio, coherence, detail, and realism. It can generate amazing images from text prompts that look almost like real photos. Some people have pointed out that artistically, v4 sometimes produces more creative results. If you’re looking for realism, however, v5 is definitely the better choice.