Midjourney V8.2 Full Guide

Updated: 2026-08-29. Default versions, parameter names, personalization options, and billing follow midjourney.com and in-product docs/announcements that day. Third-party entries ≠ official.

Introduction

Midjourney V8.2 is, in 2026 conversation, a common flagship default model (confirm what your Create page actually shows after login). It carries everyday finals for photo-real work, design-led composition, and general delivery: more willing to use current aesthetics than clinging to outdated habits, and more suitable for product and marketing visuals than forcing an anime mode. This guide is about shipping stable frames on V8.2—positioning, comparison selection, personalization, a reproducible workflow, and a delivery checklist.

What this guide solves

• Clarify V8.2’s role—and when to switch to V8.1 or Niji 7
• Build a reproducible workflow: aesthetic personalization + structured prompts
• Understand aspect, stylize strength, and redraw-class actions (names follow the UI that day)
• Use “draft for direction → lock composition → then HD” to waste fewer credits
• Catch common failures with a pre-delivery checklist

Where V8.2 sits

Treat it as the default generate tier: article heroes, marketing key visuals, product concepts, environment mood, flat layouts that need some design sense. Quality and controllability land in the everyday delivery sweet spot. Within the same generation, prefer it over clinging to outdated defaults; for thick anime paint, switch to Niji instead of cranking V8.2 parameters until it “pretends to be anime.”

Product name, on-screen label, and third-party button text may not match word-for-word. When you write specs, record: entry URL + UI-shown version + personalization snapshot + date. Identity check: What is Midjourney?.

China third-party pages may still show older labels like “V7”—fine for learning prompt structure; delivery acceptance should confirm on the official site that you are truly on V8.2. Convenience entries: Midjourney V7 · Text-to-image studio.

Choosing among V8.1, V8.2, and Niji

Dimension	V8.1	V8.2 (this guide)	Niji 7
Role	Same-gen comparison / asset alignment	Flagship everyday workhorse	Anime / illustration lean
Aesthetic center	Continues prior-tier habits	Current default all-round quality (per official)	Anime, characters, Japanese illustration
Recommended use	Regression tests, prompt migration	Everyday finals	Character sheets and anime covers
Avoid	New projects refusing upgrades with no reason	Sole answer for pure anime line art	Sole source for hard photo-real metal products

Quick decisions:

1. Need direction only and many variants → explore on a cheaper/faster mode first (mode names follow the UI), then finalize on V8.2.
2. Need a final with photo-real / design lean → use V8.2.
3. Clear anime track → open Niji 7; don’t stack “anime” adjectives on V8.2 and force it.
4. Team norms built on V8.1 samples → fixed sample-set compare before a full cutover.

Aesthetics & personalization: louder than “god prompts”

In the V8.2 era, personalization and aesthetic preference often matter more than ten extra adjectives for “does this look like your work.” Treat personalization as project config—not a one-time mysticism switch.

Lock three things before you run

1. Current model: selector shows V8.2 (or the official default flagship labeled that day).
2. Personalization state: on/off, which Profile / Moodboard (names follow the UI); switching accounts switches aesthetic distribution.
3. Default aspect: 16:9, 1:1, 9:16—set first, then write the subject, so late hard crops don’t destroy composition.

For teams, put a “personalization snapshot note” in the design spec. Otherwise A’s strongly personalized samples will never reproduce for B with personalization off.

Personalization principles

• Explore: personalization on is fine—find brand vibe fast.
• Compare: same prompt with on/off for several frames; log which side closer to acceptance.
• Deliver: lock a verified config; no random Profile hopping.
• Migrate: before a new machine or cookie clear, export/record settings the official way.

Stronger is not always better. Over-strong personalization can ignore a new product’s structure, pull every frame toward the same lighting, or refuse a flat style you explicitly asked for. When “the prompt is clear but the image ignores you,” debug personalization before blaming the model.

Build a brand sample set before you crank personalization

Before strong personalization, prepare 8–12 real delivery subjects (with aspect and bans). Generate several frames with the same prompt under off/on personalization. Score obedience, brand-color closeness, and rework time. If scores are close, pick the lower-variance side as default; if they diverge hard, document personalization as a project switch—don’t leave every designer guessing.

Re-run the sample set periodically: after official default or weight tweaks, yesterday’s “pretty” may stop reproducing. Quarterly regression on the same prompt library is cheaper than waiting for a major client complaint.

Reproducible workflow (prompt + parameter thinking)

The goal is not a novel in one prompt—it is one verified claim per round. Parameter names (stylize strength, variety, version flags, etc.) shift; below we describe intent. Concrete switches follow official docs that day.

Step 0: write the acceptance sheet (before you prompt)

Four lines in notes, then open Create:

1. Use: hero / poster KV / ecommerce main / slide illustration
2. Canvas: aspect; title reserve or not
3. Subject & count: how many objects, facing, faces allowed
4. Bans: no watermarks, no garbled glyphs, no extra limbs, no unauthorized brands

Round 1: composition only

Prompt order: use → subject → composition/camera → light & palette → material/style → on-image text (optional skip) → bans. Keep it short; don’t dump every style adjective in round one.

Use: 16:9 blog hero with right-side whitespace for a title.
Subject: one matte ceramic desktop speaker, slightly left of center.
Camera: eye level, ~50mm, soft shallow background.
Light: window soft light; cool gray and warm wood.
Style: clean product photography—no illustration feel.
Ban: watermarks, borders, extra cables, warped buttons, any on-image text.

Accept only: correct subject, correct aspect, enough whitespace, no malformed structure. Wrong composition → full reroll—don’t pile local edits on a bad crop.

Round 2: change one variable

State what stays, then the change. Use one intent among Vary / Remix / region redraw—don’t change light, material, and text in the same round.

Keep the same camera, subject pose, palette, and background.
Do one thing only: change the speaker surface from matte ceramic to brushed aluminum.
Do not change composition; do not add text or logos.

Round 3: references & consistency (when needed)

Upload one or more licensed references and say what each controls (product structure / palette / composition). Don’t ask to imitate living artists; don’t upload unlicensed assets. Product shots must be checked against source files for logos and structure.

Whether reference strength or character-lock features exist—and what they are called—follows the official site that day.

Round 4: HD and final check

Only after composition and detail pass, use upscale / HD-class actions (Upscale, etc.—UI names that day). HD cannot fix wrong composition or wrong glyphs—it only enlarges existing problems.

Zero-to-first loop: Getting started. Prompt craft: Prompt playbook. Style capture: Styles & scenes.

Draft to HD: practical credit hygiene

1. Lock aspect and use first, then spend credits on polish.
2. Draft stage: short prompts, fewer adjectives, an acceptable speed tier (UI modes that day)—direction, not pixel perfection.
3. Shortlist 1–2 frames before personalization tuning or HD.
4. Text strategy: zero-tolerance titles → “no-text image + layout tool type.”
5. Batch exploration with a fixed variable table: change only light or only camera so you can attribute.
6. Log seed / job ID if the UI provides them—rollback to a good branch.
7. Third-party draft → official final: for learning in China, Midjourney V7 or Text-to-image studio can explore structure; important delivery cross-checks on midjourney.com with V8.2.

Drafts can deliberately “write ugly”: fewer adjectives, fewer style words—subject and camera only. Pretty-but-uncontrollable frames often come from stacking mood words too early. After composition shortlists, add material, light, and brand color one layer at a time—with an acceptance check each layer. Credits and calendar usually improve.

If only one cell in a grid is barely usable, prefer directed variation on that cell over rerolling the whole set without changing the prompt—the latter replaces diagnosis with luck.

Aspect, stylize strength, and common parameter intents

Common aspects: article/video covers 16:9 (title safe zone), social square 1:1, vertical 9:16 or 3:4, slides 16:9 or 4:3.

Intent	What you’re controlling	Practice note
Aspect	Crop and safe zones	State aspect early in the prompt
Stylize strength	Artistic freedom vs prompt obedience	Too high → pretty but disobedient
Variety	How different the four cells feel	Higher while exploring; lower for finals
Version	Confirm V8.2	Don’t mix unlogged old defaults
Personalization	Aesthetic distribution	Lock for delivery; compare while exploring

Concrete spellings, defaults, and whether a flag still exists always follow official docs that day; this guide does not maintain an aging parameter table.

Minimum “prompt pack” for teams

Reusable prompts as short templates beat sharing one long prose blob:

[use / aspect]
[subject & count]
[scene & time]
[camera & light]
[3–6 style keywords—no piles]
[text: none / or exact string]
[bans]
[personalization: on/off + Profile name]
[model: V8.2 + date]

Reviews argue about empty fields, not vibes. The prompt guide expands negatives and style cards; this page’s point: V8.2 cannot rescue a brief with no acceptance criteria.

On-image text: raising the hit rate

V8.2 can draw letters; it is not a print-grade typesetting engine.

1. Keep copy short: headline-length only; legal microcopy, QR codes, dense tables rarely land correct in one pass.
2. Quote exact text and declare language: white sans title top-right: "Spring Drop".
3. No-text first, type in design tools when brand VI is strict.
4. Separate round for text changes: keep composition and subject; change only the title.
5. Final check: character-by-character against source copy; mixed scripts, digits, and ® marks fail most often.

Completion checklist

• Entry and model identity logged (UI name / date); confirmed V8.2, not an old default or Niji by mistake
• Personalization recorded or deliberately off (per project rule)
• Aspect matches use; title/crop safe zones enough
• Subject count, hands/edges, reflections and shadows without obvious collapse
• Multi-round edits changed one variable each time, with “what stays” written
• Reference purpose stated; usage rights recorded
• On-image text proofread—or switched to no-text + post type
• No surprise logos, watermarks, signatures, unauthorized brands
• Final check at target sharpness; thumbnail and full-res both reviewed
• Prices, credits, parameter defaults not frozen into external promises (official that day)
• If drafts came from a third party, cross-checked on official V8.2

Quick access

• China-friendly entry (third-party): Midjourney V7
• Mirror studio: Text-to-image studio
• Official site: www.midjourney.com

Third-party entries ≠ Midjourney’s account system; model lists are as-of that day (platforms may still show V7). Read privacy and data notes before uploading commercial assets. China access detail: China access guide.

FAQ

Is V8.2 always the official default?

In 2026 talk many people treat it as the flagship default, but your Create page that day wins. Midjourney may change defaults; this guide does not promise a permanent default.

When must I return to V8.1 instead of staying on V8.2?

When brand specs, client-signed samples, or prompt libraries are explicitly built on V8.1 look—and a sample set shows V8.2 style shift is unacceptable. Otherwise rebuild norms on V8.2 rather than locking the old tier forever.

When should I switch to Niji 7?

When acceptance criteria explicitly require anime / comic-illustration aesthetics. Stacking anime keywords on V8.2 is usually less stable than switching tracks.

Can one round demand “HD + perfect text + follow every ref”?

Not recommended. Split composition → reference constraints → text → HD so failures are diagnosable and credits stretch further. When multi-round edits keep collapsing, return to a composition-correct branch or open a new job with a good frame as reference.

What are prices and credits?

No hard-coded unit prices or quotas. See midjourney.com subscription and billing that day.

Is there an official public API to batch-run V8.2?

There is no publicly available official developer API (per Midjourney statements that day). For batch needs, evaluate compliant options or other models with public APIs—do not write third-party wrappers as “official V8.2 API.”

Official resources

• midjourney.com
• In-product docs, version notes, personalization and parameter help (whatever is visible after login that day)

Further reading

• Midjourney Guides Hub
• What is Midjourney? Family map
• Midjourney China access guide
• Midjourney Niji 7 selection & practice
• Midjourney getting started: first image
• Midjourney prompt playbook
• Midjourney styles & scenes

Summary

V8.2 is the everyday default flagship workhorse (per official default that day): lock model and personalization first, then advance composition → single-point edit → HD. Defer text when you can; run the checklist before delivery. Third-party reach ≠ official V8.2; no official public API; parameters and defaults always follow midjourney.com that day.