Best Midjourney settings for consistent e-commerce images

Recommendation
Use a single, strict prompt template + fixed MJ parameters (aspect, stylize, quality, seed strategy) and run in controlled batches from a CSV. Validate a small sample (10–20 SKUs), iterate, then scale. This gives the most consistent studio-style catalog with minimal manual curation.

Decision criteria (pick what matters most)
- Output fidelity vs cost: higher --quality and more iterations = better images but higher time/cost.
- Determinism vs variety: use seeds to make results repeatable; different seeds per SKU if you need uniqueness.
- Team/automation: small teams can use Discord + manual copy/paste; larger teams should automate with the web/API or scripting.

Prompt template (replace placeholder parts)
[PRODUCT NAME] on seamless white background, studio 3-point lighting, soft natural highlights and soft shadow beneath product, front/45° center composition, true-to-life color, accurate texture detail, no props, isolated product, photographic realism, 50mm lens, f/8, ISO 100, high detail --ar 4:5 --v 5 --stylize 50 --quality 1 --seed 123456

Notes on template
- Keep the descriptive portion identical across all SKUs except the product name/short descriptor.
- Use “seamless white background” or “neutral gray background” to force consistent backgrounds.
- Aspect ratio: 4:5 is common for e-commerce; use 1:1 for thumbnails.
- Stylize: low (e.g., --stylize 50) reduces creative variance. Adjust slightly if results look too flat.
- Quality: --quality 1 is a good balance; use higher only if needed and budget allows.

Seed usage strategy
- One seed per angle: pick a seed for your chosen composition/angle so lighting/placement is consistent across SKUs.
- Or one deterministic seed per SKU (e.g., hash of SKU) to allow repeatability while keeping each SKU unique.
- If you want identical framing for every SKU, use the same seed + same prompt and add an image reference mask (if you have product photos) to keep placement exact.

Batch workflow (practical)
1) Prepare spreadsheet with columns: SKU, PROMPT_TEXT, IMAGE_REF (optional), SEED.
2) Test: generate 10–20 diverse SKUs. Review alignment, color, shadows.
3) Tweak prompt/stylize/camera if necessary. Lock parameters once satisfied.
4) Run in batches of 25–50 images to avoid throttling; keep logs (prompt+seed+message link).
5) Post-process: auto-crop/align, batch color-check (match white point), and run a quick QA pass for 5–10% of images.
6) Re-run failures with adjusted seed or small prompt tweak.

Practical checklist
- [ ] Lock aspect ratio and camera tokens
- [ ] Choose seed strategy and document it
- [ ] Use single prompt template with placeholders
- [ ] Run small sample, approve, then bulk-generate
- [ ] Keep a CSV log of prompt+seed per SKU
- [ ] Batch post-process (crop, color, background cleanup)

Best-for / Avoid-if
- Best for: teams needing 100–500 consistent catalog images quickly, when you can accept small manual QA.
- Avoid if: you need pixel-perfect product placement identical to photography or absolute color-accurate proofing (consider studio photography or render-to-order).

When it depends
- Budget: higher MJ plan / more GPU time reduces re-runs and manual curation.
- Skill: automation and prompt tuning reduce manual work but require some scripting and QA skills.
- Team size: larger teams should automate generation + QA; solo creators can do small-batch manual checks.

Final note
Keep one canonical prompt file and a reproducible seed strategy. That single source-of-truth cuts curation work dramatically and makes bulk re-generation trivial if you change rules later.

Compare Midjourney and Leonardo AI