Midjourney V7 Arrives: Is It Better Than V6.1 and ChatGPT?

Midjourney V7 is here, but it's just the start of bigger changes coming over the next few months. The first update, version 7.1, brings some interesting new features. There's a draft mode, updates to the prompt bar, and ways to personalize your images. We also took a look at how this new version compares to the previous V6.1 and even the image feature in ChatGPT.

First Look at Midjourney V7 Features

Version 7.1 includes some noticeable shifts. The look and feel may change again as Midjourney rolls out more updates. The biggest new item right now is draft mode. This feature seems to compete directly with image tools like the one in ChatGPT.

Quick Draft Mode

Draft mode lets you get images much faster. You can even use your voice with a microphone to enter prompts, which is handy. The drawback? The images look blurry or out of focus. They are like quick ideas, not final pictures. But they are made 10 times faster than usual.

The Updated Prompt Bar

The prompt bar looks similar to what we know, but there are a few additions. You'll see a 'P' icon for personalization. There's also the option to turn draft mode on or off here. Remember, draft mode gives you early versions only. You need to refine them.

Working with Images in Midjourney

Using other images as guides has changed, especially between versions. In version 6.1, you could drag images right into the prompt bar. You could tell Midjourney if you wanted the image for inspiration, style, or character features. You could mix and match these. A tiny 'X' on the preview lets you remove images you don't need.

In version 7, you can still drag images in. The 'describe' function is there, which helps create a prompt from an image. However, the option for specific character reference seems to be missing or not yet available in V7.

Easier Personalization

Personalization, or using your saved styles (mood boards, personalization codes), is simpler in V7. You just click the 'P' icon and turn it on. It uses your chosen style automatically. Before, you had to go into settings, find your profile, and copy a long code into your prompt. The new icon is quicker and helps avoid clutter in your prompt. The only slight issue is you need to remember what each saved style name does without seeing a preview image instantly (though browser choice affects this, as we'll see later).

To streamline your creative process and manage these styles and prompts, consider using a tool like the Midjourney Automation Suite from TitanXT. It can help organize and apply your saved settings efficiently, enhancing your workflow beyond the standard interface.

Testing the Draft Mode

Let's look closer at draft mode. When you turn it on, a microphone icon appears. Clicking it lets you speak your prompt instead of typing. Midjourney detects this and switches to version 7 automatically if you aren't already using it.

Draft Mode in Action

You start with a simple idea. For example, saying "A house in Iceland" creates a basic picture. You can then add details by speaking more commands, like "Add an elephant in the foreground" or "Change the color to lush green." Midjourney updates the image concept based on your spoken words.

Once you like the overall concept, you click 'enhance'. This tells Midjourney to create four higher-quality images based on your final prompt. These enhanced images aren't exact copies of the blurry draft but follow the prompt's ideas with better detail. The elephant, car, and house were all present in the enhanced images from the examples.

Midjourney V7 vs. V6.1 vs. ChatGPT

A key part of reviewing V7 is seeing how it stacks up against its own past version and competitors. We tested seven different prompts, some simple, some quite complex, across Midjourney 6.1, Midjourney 7, and ChatGPT's image feature. The results were surprising.

The Kitten and Bear Test

Prompt: Cute kittens running from a bear on a highway. The front kitten has a "help" sign.

• Midjourney 6.1: Looked pretty good overall, with a few small issues.

• Midjourney 7: Produced very strange images, missing key elements. Showed a lot of room for improvement.

• ChatGPT: Followed the prompt perfectly. Kittens, sign, bear – everything was right where it should be.

Studio Ghibli Style Scene

Prompt: Lighthouse keeper watching a giant wave in Studio Ghibli style.

• Midjourney 6.1: Created a decent image.

• Midjourney 7: Also created an image, but didn't capture the style as well.

• ChatGPT: Delivered the winning image again, capturing the style and scene more accurately.

Minecraft Jungle Scene

Prompt: Military helicopter in Amazon jungle, dinosaurs below, white flag on helicopter saying "Which AI is better?". This prompt was very complex.

• Midjourney 6.1: Made a complex attempt but missed many details.

• Midjourney 7: Also struggled significantly with the complex prompt.

• ChatGPT: Handled the complexity much better, placing elements correctly and even attempting the text on the flag.

Country Singer Concert

Prompt: Country singer on stage at night, crowd, band, large screen behind her showing her face large.

• Midjourney 6.1: Got most elements but didn't show the face on the screen.

• Midjourney 7: Showed the face on the screen but put the singer's hat outside the screen area, which wasn't logical.

• ChatGPT: Came closest to the original prompt, showing the face on the screen correctly, even if the face rendering itself wasn't perfect.

Tomato and Toast with Text

Prompt: A tomato holding a sign saying "Let's have a big Tomaato, we'll pay all the expenses." (This prompt had a deliberate typo). Also involved charred toast.

• Midjourney 6.1: Created a good photorealistic image, but the text on the sign was wrong or unreadable.

• Midjourney 7: Created a strange, almost unrecognizable image. The text was also wrong and nonsensical.

• ChatGPT: Produced an illustration (like V7, while V6.1 was photo-like) but nailed the text on the sign perfectly, including the typo. It followed the text instruction far better.

Nested Items

Prompt: A bottle in a bucket. Bucket on a ladder. Ladder against a box. Smoke from box looks like a skull.

• Midjourney 6.1: Created elements but placed them next to each other, not nested. Smoke came from the bottle, not the box, and didn't look like a skull.

• Midjourney 7: Got the nesting correct (bottle in bucket, bucket on ladder) but the ladder wasn't against the box. Smoke came from the box and looked slightly like a skull, but the overall scene didn't match the prompt exactly.

• ChatGPT: Arranged all elements correctly according to the nested sequence (ladder, bucket, bottle, box with skull smoke).

T-Shirt Mockup with Curved Text

Prompt: Vector style kitten on yellow t-shirt. "Cute fellow" in curved font above kitten.

• Midjourney 6.1: Text was below the kitten, not above.

• Midjourney 7: Looked visually nice but had strange details at the kitten's feet and missed a letter in the text ("fellow").

• ChatGPT: Followed the prompt perfectly. Kitten (black as requested), text above, curved, and spelled correctly.

Based on these tests, Midjourney V7 seems to struggle with detailed prompt adherence, especially text, more than V6.1 and significantly more than ChatGPT. It felt unfinished in key areas where ChatGPT excelled.

Managing prompts and tracking results across different tests can be time-consuming. Automate parts of this process and organize your results by exploring the Midjourney Automation Suite from TitanXT.

Finding Your Images with Organize

The 'Organize' section within Midjourney helps you manage your creations. While not brand new to V7 necessarily, understanding it is useful.

View Options

You can change how your images display. The default might be large, requiring lots of scrolling if you have many images. Changing the view to 'small' shows many more images at once, making it easier to quickly find something by looking at the pictures. 'Medium' is also an option, but 'small' offers the best overview for browsing many results.

Sorting by Style

In your profile, you can sort images by your saved personalization codes or mood boards. Clicking a style name shows only the images you created using that specific style. This is very helpful if you use many different personalization codes.

Saved Searches (Filters)

You can create your own search filters. You define keywords (like "monster" or "animal, world"). Then, clicking that saved search option quickly shows all images whose prompts included those words. To create one, hit 'add search', give it a title, enter keywords separated by commas, and save.

Filters

On the side, there are general filters. You can sort by rating, type (like image size), version number (V6.1 vs V7), publish status, and more. You can use multiple filters together to narrow down results.

Does Your Browser Matter?

Using Midjourney V7, your choice of browser seems to make a difference. In comparisons between Firefox and Google Chrome, key features worked better in Chrome.

For example, in the personalization section, Chrome displayed small preview images for each saved style (mood board). This lets you see what a style looks like quickly. In Firefox, these preview images were missing. You had to click each style name to see the images created with it, which is less efficient.

More importantly, the new microphone feature for voice prompts in draft mode did not seem to work in Firefox. The icon was there, and settings looked correct, but clicking it did nothing. In Google Chrome, clicking the microphone immediately activated voice input. If you plan to use voice prompting or rely on visual previews for personalization, Chrome worked better in these tests.

Final Thoughts on Midjourney V7

Midjourney V7 feels less finished than expected, especially when looking at the image comparison tests. Where V6.1 often did reasonably well and ChatGPT was very accurate, V7 sometimes produced strange results, particularly with text and following complex instructions precisely.

The draft mode requires a different way of thinking. You start with a basic idea, refine it step-by-step using text or voice, and only then use 'enhance' to get better quality images. This is different from just writing one detailed prompt and expecting a perfect result immediately, which is closer to how ChatGPT's image tool works.

While the ability to refine is exciting, if you know exactly what you want from the start, the ChatGPT approach feels simpler and more reliable based on these tests, especially for things like adding text accurately. Text in images is still a point where Midjourney V7 struggles significantly, which can be a real drawback for many uses.

It will be interesting to see how Midjourney develops V7 with future updates. They mention features like 3D and video might be coming. For now, using tools that can help bridge some gaps, like organizing your prompts or quickly iterating through ideas, can be valuable. Check out the Midjourney Automation Suite by TitanXT to improve your workflow.

Version 7 is a work in progress. It brings speed with draft mode and easier personalization access, but it has areas that need improvement, especially in following detailed prompts compared to its predecessor and competitors.