
Exploring Midjourney V7: First Impressions from 50 Prompt Tests
May 3, 2025
5 min read
0
1
0

Midjourney's version 7 is here, bringing new ways to create AI art. This latest update changes how the model interprets prompts and introduces some features. Let's look at how V7 performs with various prompts, from simple objects to complex scenes and abstract ideas, and compare the default results with a personalized style.
Getting Started with V7
To use V7, you might need to unlock personalization on the Midjourney website. If you had personalization for V6, it seems V7 requires a separate unlock process. You might even need to rank images on the website before you can fully use V7. This personalized style is different for each user based on their ranking choices.
New Features in V7
V7 includes a new Draft Mode. This mode is designed to be faster for iterating on ideas. Midjourney also mentioned a conversational aspect, where you can speak into a microphone to help generate prompts. This could be a unique way to brainstorm ideas before refining your text prompts.
For those looking to enhance their Midjourney workflow and potentially automate some tasks like prompt generation, testing, and organization, consider the TitanXT Midjourney Automation Suite. It can help you manage your creative process more effectively.
Speeds and Other Models
Currently, V7 offers Turbo (faster, uses more hours) and Relax modes (slower, no hour usage if available). The standard Fast speed is not available yet. Features like upscaling, inpainting, and retexture still use V6 models, even if the initial image was created with V7.
Testing Prompts: Default vs. Personalized Style
Testing Midjourney often involves seeing how it handles different subjects. Here's what V7 produced with a range of prompts, comparing the default model with a personalized style.
Simple Subjects & Objects
Sphere: Results leaned towards photographic and slightly surreal.
Woman / Man / Dog / Cat: Default style often produced close-up, photographic images. The personalized style sometimes showed different tendencies, like colder tones for animals or more rugged looks for men.
Architecture and Nature
Castle: Default gave a mix of familiar castles, ruins, and painted looks. Personalized style seemed to reduce clutter and detail.
Forest: Default created realistic forest scenes. Personalized style added effects like fog or a turquoise tint.
Meadow: Default results were surprisingly often painted rather than photographic.
Lighthouse: Default had a painterly feel with toned-down colors. Personalized style was more photographic or minimalistic.
Keeping track of prompt variations and their results can be time-consuming. Tools like the TitanXT Midjourney Automation Suite can assist in managing prompt tests and comparing outputs efficiently.
Sci-Fi and Fantasy Themes
Spaceship: Provided a wide variety of sci-fi styles. Personalized style strongly reflected the user's preferred sci-fi aesthetics.
Vampire: Default was photographic and focused on teeth. Neither style consistently depicted vampire fangs well.
Mermaid: Still appears difficult for V7 to generate successfully in either style. Results often looked like dresses rather than tails.
Centaur: V7 struggled with this mythological creature, mostly showing people on horses instead of actual centaurs.
Detailed Objects and Food
Violin: Default was realistic but not perfect. Personalized style surprisingly had more color.
Typewriter: Looked correct from afar, but details like letters were incorrect upon closer inspection.
Croissant: Default was realistic with no plates, matching the simple prompt. Personalized style added plates and illustrations, looking less like actual croissants.
Backpack: Default produced vintage-looking photographic designs. Personalized style gave more expected, modern photographic results.
Robot: Default created cute-looking robots. Personalized style kept the cute look but with more subtle colors or illustrations.
Android: Both styles fixated on the Android logo rather than creating android characters.
Maximizing your creative output with Midjourney V7 might require extensive testing. Streamline this process with the TitanXT Midjourney Automation Suite, built to help you test prompts and manage your creations.
People and Crowds
Realistic Hands: Default showed five fingers with some minor unnaturalness. Personalized style had stranger results.
Child reaching for butterfly: Some images were as expected, but others were odd or focused only on the hand.
Cybernetic man with different eyes: Neither style successfully showed one human and one mechanical eye; only one eye was visible.
Two people shaking hands: Default looked okay, but skin tones for both hands in a single image were always the same. Personalized style moved the scene outdoors.
Crowded subway / Choir singing: V7 still seems to avoid clear, visible faces in crowd scenes. Faces are often obscured, far away, or turned away.
Period Looks (Flapper girl, Renaissance woman): Default and personalized styles tended to look like modern photos of people in costume rather than period-appropriate images. Faces were sometimes cut off in wider aspect ratios.
Long shot of a woman: V7 did well showing the full figure, though the face wasn't clear.
Text Prompts
Fantasy book cover with title: Success rate for text was good, producing several nice results from both styles on the first try.
Wanted poster with text: Had more difficulty with text accuracy compared to the book cover prompt.
Impossible Concepts
Ballerina made of flowing ink: Default showed ink drawings. Personalized style had better results, including one image closer to the desired concept.
Fox made of candle wax: Created fox-shaped candles that weren't melting much.
Staircase made of moonlight: Did not grasp the concept well, showing moons by staircases.
Tree made of music notes: Looked like paper with notes on it or cut out from it.
Bridge made of spider silk: Produced interesting but abstract interpretations.
Abstract Ideas and Art Styles
Something: Default was less colorful than previous versions, often with a different painterly look. Personalized style was described as epic and weird.
Equilibrium: Default leaned towards minimalism. Personalized style mixed minimalist photography with detailed illustrations.
Void: Produced nice, atmospheric photos or illustrations.
Harmony: Default was mostly painterly. Personalized style created gorgeous images.
Whimsical Surrealism: Default had more detail but muted colors compared to earlier versions. Personalized style also muted colors.
Coloring book page: Became too detailed in V7 to be useful as a coloring page template.
Zenangle: Default was too detailed. Personalized style was also too detailed for a Zenangle purpose.
Fashionable alien: Results were nice but considered less cool than previous versions.
Grotesque beauty: Leaned more grotesque than beautiful in V7. Showed unexpected ornate details.
Art Nouveau: Default was spot-on. Personalized style was loved for its color tones and shapes.
Pointillism: Default results were nice, some abstract. Personalized style favored turquoise colors and circles made of points.
Overall Impressions of V7
Midjourney V7's default style is noticeably different from V6, with a changed overall vibe. It handles single people and text (though not perfectly) better than before. However, it still struggles with complex concepts like mythological creatures and depicting multiple clear faces in a single image.
The personalized style in V7, which starts fresh, seems to work very well for users, closely matching current taste preferences compared to older versions influenced by long ranking history.
Conclusion
Midjourney V7 offers a fresh experience with a distinct default style and seemingly improved personalization. While it excels in some areas like single subjects and text, challenges remain with complex scenes and specific creature details. Experimentation is key to mastering this new model.
Continue exploring Midjourney V7 with tools designed to make your process smoother. The TitanXT Midjourney Automation Suite can help you manage your prompts and tests as you discover what V7 can do.






