Midjourney v6 vs. Flux: The Ultimate Typography & Camera Lens Prompting Guide

Two engines, two philosophies. Midjourney v6 wants to be art-directed; Flux wants to be engineered. Master both syntaxes and you can produce billboard-quality visuals on demand — including the one thing that broke every prior generation: legible, on-brand typography.

1 · Typography: The Quote-and-Anchor Method

Both engines now render text — but only if you constrain them properly. The single biggest unlock is wrapping your text in straight double quotes and following it with an explicit anchor describing where the text sits and what material it's rendered in.

Midjourney v6:

luxury whiskey bottle on charcoal velvet, label reads "AURUM 1888"
in embossed gold foil serif type, centered on the label, raking
side light, --ar 4:5 --style raw --stylize 200 --v 6.1

Flux (Pro / 1.1):

A luxury whiskey bottle on charcoal velvet. The label displays the
text "AURUM 1888" in embossed gold foil serif typography, perfectly
centered, with crisp letterforms and visible foil specularity. Raking
side light from camera-left. Studio product photography. 4:5 aspect.

Three rules that hold across both engines:

• ≤ 4 words per text block. Beyond that, both engines hallucinate glyphs.
• Name the material (foil, neon, chalk, letterpress). The material constrains the renderer's typography model.
• Anchor the position ("centered on the label", "top-left corner"). Floating text drifts into gibberish.

2 · Camera Lens Prompting — Exact Parameters

Stop saying "cinematic." Specify the actual glass. Both engines were trained on EXIF-tagged datasets; the lens spec triggers the right look almost deterministically.

// Portrait, creamy bokeh
shot on Sony A7R V, 85mm f/1.4 GM, ISO 200, 1/250s, eye-level,
shallow depth of field, background dissolves into circular bokeh

// Architectural, deep focus
shot on Hasselblad X2D, 38mm f/8, ISO 64, 1/60s, hyperfocal,
edge-to-edge sharpness, no distortion

// Editorial fashion, compressed perspective
shot on Mamiya RZ67, 110mm f/2.8, ISO 100, medium format film grain,
shallow plane, soft falloff

The non-obvious lever is aperture as an emotional dial:

• f/1.2 – f/1.8 — intimate, dreamlike, isolates subject. Use for portraits and product hero shots.
• f/4 – f/5.6 — editorial, balanced, professional documentary feel.
• f/8 – f/16 — architectural, journalistic, "everything matters." Use for environments and group scenes.

3 · Midjourney-Specific Parameter Cheat Sheet

--ar 4:5        // editorial portrait (Instagram-native)
--ar 16:9       // cinematic landscape
--ar 21:9       // ultrawide / banner
--style raw     // strips MJ's house aesthetic — critical for realism
--stylize 50    // low: literal interpretation of prompt
--stylize 250   // default: balanced
--stylize 750   // high: MJ takes creative liberties (use for art, not product)
--chaos 0       // consistent variations across the 4-grid
--chaos 50      // wild variations — use for ideation, not finals
--weird 0       // grounded
--weird 250     // surreal artifacts encouraged
--seed 12345    // reproducibility — pin this once you find a good roll

4 · Flux-Specific Discipline: The Three-Paragraph Structure

Flux rewards prose; Midjourney rewards keywords. Structure every Flux prompt as exactly three paragraphs:

1. Subject paragraph — who/what, materials, expression, pose.
2. Environment paragraph — setting, time of day, weather, scale relationships.
3. Camera paragraph — body, lens, aperture, ISO, shutter, post-processing, aspect ratio.

Mixing these collapses Flux's attention. Keep them separate and the engine treats each as an independent rendering constraint.

5 · Contrast Rules That Make Images Pop

Visual prompts live or die on contrast. The four contrast axes — and how to invoke them — are:

• Tonal contrast → "deep shadows, blown highlights, no mid-tones" (chiaroscuro)
• Color contrast → "complementary palette, teal and orange grading"
• Texture contrast → "smooth marble against rough burlap"
• Scale contrast → "single figure dwarfed by colossal architecture"

Invoke at least two contrast axes per prompt. Images with only one feel flat regardless of which engine generated them.

The era of "pretty pictures from vague prompts" is over. The era of art-direction-as-code has begun. Engineers who learn the syntax will out-produce illustrators who don't.