Midjourney V7 Guide: New Features, Prompting Tips & What Changed

What Is New in Midjourney V7

Midjourney V7 represents the most significant generational leap since V5 introduced photorealism to the platform. The model brings improvements across photorealism, text rendering, prompt coherence, and a new personalization system that learns your aesthetic preferences over time. If you have been using Midjourney for a while, some prompts that worked well in V6 may need adjustment — but the ceiling on quality is substantially higher.

This guide covers the key V7 changes, how to adapt your prompting strategy, and which new features are worth learning first.

Key Improvements in V7

Enhanced Photorealism

V7's photorealism improvements are most visible in portrait photography and architectural visualization. Skin textures, fabric detail, and environmental lighting are rendered with significantly greater accuracy. Complex lighting scenarios — dappled sunlight, neon-lit scenes, candlelight interiors — that previously required multiple Remix iterations now often render correctly on the first generation. For commercial photography use cases, V7 narrows the gap between AI generation and actual photography.

Improved Text Rendering

Text in images has historically been a weakness across all AI image generators. V7 makes meaningful progress here — short text phrases in logos, signs, and labels render legibly in many cases. For simple single-word or short-phrase text integrated into compositions, results are substantially more usable than previous versions.

Prompt Coherence and Instruction Following

V7 demonstrates better adherence to complex compositional instructions. Specifying a subject placement, lighting direction, or specific camera angle now yields results that more consistently match the description. Multiple subjects with different characteristics are better differentiated. Negative prompting with --no is more effective at genuinely excluding unwanted elements.

Personalization System (--p Flag)

The personalization feature, accessed with the --p flag, is V7's most distinctive addition. By rating hundreds of images in the Midjourney web app, you train a personalization profile that learns your aesthetic preferences. Once established, appending --p to any prompt biases generation toward your preferred style — more vivid or muted, more abstract or literal, more painterly or photographic. The more images you rate, the more precisely the system captures your taste.

Prompting Strategies for V7

Be More Specific About Technical Parameters

V7 responds better to technical photography language than previous versions. Specify lens length (85mm portrait lens), aperture (f/2.8 bokeh), lighting setup (softbox key light, rim light), and camera style (medium format film). These terms anchor the model's understanding of the intended rendering style and produce more consistent, professional-looking results.

Use Style References Effectively

The --sref (style reference) parameter lets you provide an image URL as a style anchor. V7 extracts the visual language of the reference — color palette, texture, compositional approach — and applies it to new generations while following your text prompt. This is invaluable for maintaining brand consistency across a series of images.

Aspect Ratios and Canvas Sizes

V7 handles unusual aspect ratios more gracefully than previous versions. For cinematic compositions, --ar 2.39:1 produces genuine widescreen framing. For mobile-first content, --ar 9:16 generates vertically composed images that work as phone wallpapers or Stories. Always specify the intended output format early in your prompt workflow.

Character Reference (--cref)

The character reference parameter maintains consistent character appearance across multiple generations — essential for illustration projects, brand mascots, or any work requiring a recognizable recurring character. Provide a reference image URL and the model preserves facial features, hair, and general appearance while adapting pose, setting, and style as directed.

V7 vs V6: Should You Switch?

For photorealistic work, commercial product imagery, and architectural visualization, V7 is clearly superior. For stylized illustration and anime aesthetics, V6 Niji may still be preferable depending on your style targets. Run parallel generations in both versions with your most-used prompts to calibrate your own judgment — different creative workflows will have different optimal model choices.

Getting Started With V7

V7 is the default model for new Midjourney subscribers in 2026. Existing users can switch by typing /settings in the Discord bot and selecting V7, or by appending --v 7 to any prompt. Start by adapting two or three of your best-performing V6 prompts to V7 to feel the differences before building new workflows from scratch. Browse our AI image generator comparisons to see how Midjourney V7 stacks up against alternatives like Stable Diffusion, DALL-E 3, and Flux.