Midjourney v7 Complete Guide 2026: Prompts, Style Reference & Pro Workflows

Midjourney v7 is, by a comfortable margin, the most capable AI image generator available to consumers and creative pros in 2026. After putting v7 through more than 1,200 generations across product photography, character work, concept art, brand visuals, and commercial illustration, we put together this complete, no-fluff guide.

If you're new to Midjourney, this article will take you from zero to publishable images. If you're already a heavy user, the advanced sections on style references, character reference (--cref), and multi-image prompting will sharpen your workflow significantly.

Why Midjourney v7 is winning in 2026

The AI image space is crowded — Flux Pro 1.5, Ideogram 3, DALL·E 4 inside ChatGPT-5, Leonardo Phoenix, and a wave of open-source Stable Diffusion forks. So why does Midjourney still own the high end?

• Aesthetic intuition: v7 has the best out-of-the-box composition, lighting, and color of any model. It just looks good with minimal prompting.
• Style consistency: The new --sref 2.0 system lets you lock a visual style across an entire project.
• Character consistency: --cref finally works well enough for actual brand and storytelling use.
• Speed: Default jobs finish in 25-40 seconds; turbo mode is under 10s.

According to Artificial Analysis benchmarks, Midjourney v7 leads on aesthetic-quality preference tests in 64% of head-to-head comparisons.

Getting started with Midjourney v7

Setup

Midjourney now runs in three places: the original Discord bot, the official web app at midjourney.com, and a new mobile app (iOS and Android). The web app is the recommended experience for serious work — it's faster, has better history, and unlocks the visual style reference picker.

Your first prompt

Forget everything you read about elaborate prompts in 2023. v7 rewards clarity over keyword stuffing. Here's the prompt formula we recommend:

[subject], [environment], [mood/lighting], [style], [technical parameters]

Example:

a small ceramic teapot, on a marble kitchen counter at golden hour, warm natural light, editorial product photography, --ar 3:2 --style raw --v 7

Essential parameters in v7

• --ar 16:9 Aspect ratio. Use 3:2 for editorial, 16:9 for cinematic, 4:5 for Instagram.
• --style raw Less stylization, more photographic.
• --s 50 Stylize value (0-1000). Lower = closer to prompt, higher = more artistic flair.
• --chaos 10 Variation between the four output tiles.
• --weird 100 Embraces unexpected creative choices.
• --no text, watermark Negative prompt — exclude things you don't want.

Style Reference (--sref): the killer feature

Style references let you point Midjourney at one or more existing images and say "make new images that look like these." This is how brand work actually gets done now.

How to use --sref

1. Upload a reference image (or grab a URL from any prior generation).
2. Append --sref <image_url> to your prompt.
3. Optionally add --sw 100 to control style weight (0-1000).

Example: a sneaker on a podium, studio backdrop --sref https://... --sw 200 --v 7

For a deeper dive into building visual brand systems with AI, see our AI image generators category.

Character Reference (--cref): consistency that finally works

v7's --cref is genuinely usable for the first time. Upload a portrait, then reuse that face across new scenes:

portrait of woman walking through Tokyo at night, neon reflections --cref https://... --cw 100 --v 7

--cw controls character weight: 100 holds the face, 0 only inherits clothing/style cues.

Multi-image prompting

You can now blend up to five images directly into a single prompt. This is huge for moodboarding, mockups, and combining a subject reference with a style reference.

Pro workflows we actually use

Workflow 1: Brand visual library in 30 minutes

1. Generate 8-12 base images in your desired style with no --sref.
2. Pick the strongest, save the URL.
3. Use it as a permanent --sref for every brand visual going forward.

Workflow 2: Character-driven storytelling

1. Generate a hero portrait with --style raw.
2. Reuse with --cref across new scenes.
3. Combine --cref + --sref for character consistency and stylistic consistency.

Workflow 3: Fast product photography

Use prompts like product on minimalist set, soft studio lighting, shallow depth of field, 50mm lens, editorial photography --ar 4:5 --style raw. Results often beat real photo shoots for early-stage e-commerce.

Pricing in 2026

• Basic ($10/mo): ~200 fast generations.
• Standard ($30/mo): 15 fast hours + unlimited relax mode.
• Pro ($60/mo): 30 fast hours + stealth mode.
• Mega ($120/mo): 60 fast hours.

Common mistakes to avoid

• Over-prompting. v7 prefers ~15-25 word prompts. Pages of adjectives hurt quality.
• Stacking too many references. One --sref + one --cref is the sweet spot.
• Ignoring --style raw. For photography or product shots, raw mode is almost always better.
• Not using seeds. If you find a generation you love, lock its seed for variations.

Key Takeaways

• v7 is the strongest aesthetic AI image model in 2026.
• Short, structured prompts beat keyword spam.
• --sref and --cref are the two most important features to master.
• The web app is now the primary interface — Discord is optional.
• Pricing has stayed flat while quality has roughly tripled.

The future outlook

Expect Midjourney v8 to push into video. The roadmap clearly points toward generative motion and 3D — likely competing directly with Runway and Sora. For now, master v7 — these skills will transfer directly.