midjourney
Midjourney V8.2 Default Model: Migration Checklist for V7 Users
MidassAI Team · September 1, 2026 · 3 min read
Keywords: midjourney, midjourney-v8, migration, personalization, draft-mode
Published: September 1, 2026 Author: MidassAI Team
If your muscle memory stops at V7 --oref and --cref, the July 2026 default switch to V8.2 will silently change client deliverables. New prompts already run V8.2 unless you force an older version.
What changed when V8.2 became default
Midjourney positions V8.2 as an aesthetics-and-personalization release—not a raw capability leap. In daily work that means:
- Outputs skew bolder and more stylized than V8.1’s flatter look
- Personalization profiles read rating history more accurately when your pool is large
- Native 2K output and faster V8-series generation carry forward from earlier V8 builds
- Image prompting and an updated
/describereturn longer reverse-engineered prompts
You no longer need --preview flags from the V8.2 preview period—every new job uses V8.2 by default.
Migration checklist (printable)
| Step | Action | Why |
|---|---|---|
| 1 | Re-rate 20–40 pairs if personalization feels stale | V8.2 personalization weights changed |
| 2 | Archive hero --sref codes that looked good on V8.1 |
Style obedience improved but defaults shifted |
| 3 | Run three golden prompts with --v 8.1 and V8.2 side by side |
Document client-visible drift |
| 4 | Move exploration to Draft; finalize without Draft | Saves GPU on experiments |
| 5 | Update prompt snippets: drop deprecated V7-only flags | --cref era is over |
Prompt habits that still work
The one sentence + two parameters pattern (--ar, --raw) remains valid. What changed is how aggressively the model interprets stylize defaults—try --stylize 500–800 on editorial work if out-of-box grids feel too loud.
Pair --profile with campaign --sref for recurring brands: profile = your taste, sref = this job’s art direction.
When to force --v 8.1
Keep --v 8.1 in pocket for:
- A/B tests during a rebrand approval
- Recreating an approved V8.1 hero exactly
- Debugging whether a failure is model-specific or prompt-specific
Do not leave --v 8.1 in production templates indefinitely—security patches and edit tooling target the default model first.
Draft mode + personalization together
Workflow that survived our internal QA:
- Draft-roll composition with personalization on
- Note winning seed and exact prompt line
- Re-run winner without Draft for delivery resolution
- Upscale only after color and crop sign-off
Red flags after migration
- Faces match but wardrobe colors drift → lower
--owon references or describe clothing explicitly - Background clones from reference boards → crop references tighter
- Client says images look over-filtered → drop stylize, add
--raw, or temporarily force V8.1 for that batch
Bottom line
Migration is not relearning Midjourney—it is re-baselining three golden prompts, refreshing personalization, and tagging which jobs still need --v 8.1. Do that once per account before the next campaign goes out.