Midjourney

Midjourney AI Generator

Midjourney is an independent research lab's text-to-image platform that transforms natural language prompts into high-quality visual artwork. Known for its distinctive aesthetic sensibility, Midjourney excels at creating breathtaking, art-directed visuals with exceptional mood, atmosphere, and compositional harmony.

Unlike prompt-literal generators that prioritize instruction-following, Midjourney interprets your descriptions through an artistic lens—often producing results that feel more like collaboration with a skilled digital artist than simple command execution.

Current Version on Somake: V7 — You can access legacy versions via the left-hand panel.

V8 Alpha Now Live: Midjourney has released V8 in alpha at alpha.midjourney.com for all subscribers. Somake integration coming soon.

Midjourney V8: What's New?

Midjourney V8 is currently in alpha testing. Here's what early access reveals:

What's Improved

• Faster Generation: Noticeably quicker image output compared to V7

• 2K HD Mode: New native 2K resolution option for high-fidelity exports

• Enhanced Photorealism: Realistic styles have improved, delivering more convincing skin textures and lighting

• Rich Creativity: V8 continues Midjourney's strength in artistic, imaginative outputs

What Still Needs Work

• Hand Accuracy: Anatomical issues with hands persist

• Text Rendering: Short text quality has improved slightly but remains mediocre

• Illustration Quality: Non-photorealistic styles (watercolor, flat illustration) have regressed

• Prompt Comprehension: Semantic understanding still trails behind competitors like Nano Banana

Should You Use V8?

For photorealistic work requiring speed and high resolution, V8 shows promise. For illustration, character consistency, or text-heavy designs, V7 or Nano Banana may still be the better choice.

What Makes Midjourney V7 Superior?

Prompt Guide

Core Parameters

Version

• 7: Most advanced stable model with rebuilt architecture

• 6.1: Optimized V6—faster processing with improved textures

• 6: Legacy stable release, useful if newer versions alter your preferred aesthetic

Niji (Anime and Illustration)

• 6: Latest iteration, excels at complex lighting, sharp details, and text rendering

• 5: Classic version with more stylized, retro anime aesthetic

• 0: Base iteration for experimental early Niji styles

Quality

• 0.5: Draft quality for fast idea testing

• 1: Standard balance of detail and generation speed

• 2: High fidelity for maximum detail on final exports

Creative Controls

Stylize (0–1000)

Controls how much creative license the AI takes with your prompt.

• Low values (0–100): Strictly follows your description—best for logos or specific technical designs

• High values (200–1000): Adds creative interpretation, enhanced lighting, and compositional flourishes

Chaos (0–100)

Controls variety among your initial grid results.

• Low values: Predictable, similar outputs across the grid

• High values: Wildly different compositions—useful when exploring creative directions

Weird (0–3000)

Controls unconventional and surreal elements.

• Low values: Standard, conventional aesthetics

• High values: Surreal, unexpected results—ideal for avant-garde or experimental projects

Style Reference Template

[subject description] --sref [reference code or URL] --sw [0-100]

Example:

a cute girl in a garden --sref 123456789 --sw 50

The --sw parameter controls style weight—higher values apply the reference more strongly.

Midjourney vs. Nano Banana

Use Cases

Version History

Troubleshooting Common Issues

Problem: The AI ignores my hex color codes.
Solution: Place hex codes at the very end of your prompt and combine them with a Style Reference image. Hex codes work best as reinforcement rather than primary direction.

Problem: Hands and fingers look distorted or have extra digits.
Solution: This remains a known limitation. Try adding "anatomically correct hands" to your prompt, use lower chaos values, or plan to fix hands in post-processing. V8 alpha has not fully resolved this issue.

Problem: Text in images is illegible or misspelled.
Solution: Midjourney's text rendering is limited. For text-heavy designs, consider generating the image without text and adding typography in a design tool afterward. Alternatively, check our guide to Midjourney alternatives for models with stronger text capabilities.

Problem: Results look too similar across multiple generations.
Solution: Increase your chaos value to introduce more variation. Start with chaos 25-50 for moderate variety, or go higher for dramatically different options.

Problem: The model keeps adding unwanted elements to my composition.
Solution: Use the --no parameter followed by elements you want excluded. Example: --no trees, clouds, people

Problem: My style reference isn't being applied consistently.
Solution: Increase the style weight parameter (--sw) to values between 50-100 for stronger application. Ensure your reference image clearly represents the style you want.