Image AI March 23, 2026

Midjourney V7 Prompting Guide: Natural Language, Parameters, and Style Control

Midjourney V7 is the current default model, and it processes prompts completely differently from V5 and V6. It shifted to natural language — full sentences written like photography briefs, not keyword lists. Most people haven't adjusted. Here's everything you need to know.

The biggest change: natural language over keywords

V5 and V6 responded well to comma-separated keyword lists. Tags like cinematic, golden hour, bokeh, 8K carried real weight. V7 still understands those terms but it now processes full sentence descriptions more effectively.

This is because V7 uses a different text encoder that parses language more like a reading comprehension model than a keyword-matching engine. It reads for intent, not just tags.

V5/V6 style (still works, but suboptimal in V7)

"woman portrait, golden hour, bokeh, soft light, film grain, Canon, 85mm, emotional, cinematic, --v 7"

V7 native style

"A woman in her late 20s photographed at golden hour, face turned slightly away from camera, soft rim lighting from the setting sun, shot on a Canon EF 85mm f/1.4 lens with shallow depth of field, subtle film grain from Kodak Portra 400, melancholic and introspective mood. --ar 4:5 --v 7 --style raw"

Word order matters in V7

V7 weights the beginning of the prompt more heavily than the end. Lead with what matters most. If your subject is a portrait, start with the person. If it's a landscape, start with the scene.

Don't bury the main subject in the middle of the prompt surrounded by style descriptors. Those should come at the end, before the parameters.

Key parameters for V7

Parameter	What it does	Notes
--v 7	Forces V7 model	Default as of early 2026
--style raw	Reduces Midjourney's aesthetic interpretation	Use when you want literal execution of the prompt
--ar 16:9	Aspect ratio	16:9, 4:5, 1:1, 9:16, 3:2 all supported
--s 0–1000	Stylization strength	Low = literal, High = artistic interpretation
--chaos 0–100	Variation between results	Higher = more unpredictable outputs
--q .5 / 1 / 2	Quality / generation time	--q 2 for final renders
--no [element]	Negative prompt	--no text, --no watermark, --no people

Photography references work exceptionally well in V7

V7 was trained on a massive volume of photography and understands camera language deeply. These references produce consistent results:

Camera bodies: Canon 5D, Sony A7IV, Leica M11, Fujifilm X-T5, Hasselblad 500C
Lenses: 85mm f/1.4 portrait, 24mm wide-angle, 200mm telephoto, macro
Film stocks: Kodak Portra 400, Kodak Ektar 100, Fujifilm Velvia 50, Ilford HP5 (black and white)
Lighting: Always name the source. Neon signs, candlelight, golden hour backlight, overcast diffused, ring light, strobe flash

Pro tip: "Dramatic lighting" tells V7 nothing. "Hard side lighting from camera left, deep shadows on right, single tungsten source" tells V7 exactly what to build.

Art direction references

V7 responds well to artistic references — photographers, filmmakers, painters — when used as stylistic direction rather than subject matter. Use them at the end of the prompt as context, not description:

"...in the style of Gregory Crewdson, cinematic staging, artificial suburban lighting"

"...reminiscent of Saul Leiter's street photography, layered depth, color abstraction"

A complete V7 prompt: before and after

What most people type

"abandoned factory interior, moody, dark, light coming through windows"

Full V7 prompt

"Interior of an abandoned industrial factory, massive rusted machinery visible in the background. Shafts of dusty afternoon light cut through broken skylights, illuminating particulate in the air. Concrete floors stained with rust and oil. Shot on a Canon EF 24mm f/2.8 at ground level, natural overcast exterior light as the dominant source, deep shadow areas with rich detail retention, environmental storytelling aesthetic. --ar 16:9 --v 7 --style raw --q 2 --no people"

When to use --style raw vs. default

The default V7 style applies Midjourney's aesthetic interpretation — it makes images look "Midjourney." For most creative work this is fine. --style raw turns this off and executes your prompt more literally. Use it when:

You want photorealistic results that don't look AI-generated
You're matching a specific reference or brief
The default style is overriding your intended aesthetic

Get perfect Midjourney V7 prompts automatically

Type your rough idea. HonePrompt writes the full V7-optimized prompt including parameters, lighting, and style direction.

Try it free