Short answer / recommendation
Use a single “anchor” composition (either an uploaded template photo or an exact camera-and-light phrasing) + the same Midjourney flags (aspect ratio, seed, stylize, quality). For strict catalog consistency, upload one perfect shot of a blank product angle and reference it in each prompt; if uploads aren’t possible, use the exact camera-angle phrasing and the same --seed and low --stylize. Use Midjourney’s upscale on the selected result and, if you need print-level pixels, re-upscale externally (Topaz/Gigapixel) after download.
Core prompt template (works as a single-line prompt)
[IMAGE_URL/] — centered, isolated on seamless pure white background, softbox studio lighting, subtle natural shadow under product, product fills ~60% of frame, front 3/4 view (camera 45° to front), eye-level camera, 90mm lens look, high-detail, color-accurate, no props/no text/no logo --ar 1:1 --v 5.2 --q 2 --s 30 --seed 12345678
How to use the template (practical tips)
- If you can upload: create one anchor image (your ideal shot: white background, exact angle/scale). Put that image first in the prompt as IMAGE_URL. That locks framing far more reliably than seed alone. Then append the (e.g., “sleek stainless-steel travel mug, 16oz, matte finish, red”).
- If you can’t upload: remove IMAGE_URL and keep the camera phrasing EXACT and use the same --seed for every SKU to reduce composition variance.
- For tall products use --ar 4:5; for wide or square-packed catalogs use --ar 1:1 (most marketplaces prefer 1:1).
- Vary only the leading product description to generate each SKU (color, pattern, material) and keep the rest identical.
Camera-angle phrasing examples (pick one and re-use verbatim)
- “front three-quarter view, camera 45° to product, eye-level”
- “straight-on eye-level, centered”
- “overhead 90° top-down, centered”
Include lens look: “90mm lens look” for compressed perspective (product photography feel).
Seed and stylize notes
- Use a fixed seed: --seed 12345678 (any integer). Same seed + same anchor/composition gives more consistent framing across different product prompts.
- Use low stylize: --s 30 (or --stylize 0 if you want minimal artistic liberty) to reduce MJ’s creative drift.
- Use --q 2 for higher detail but it costs more GPU time/credits.
Upscaling tips
- Use Midjourney’s U buttons to upscale your chosen variant (or --upbeta / --uplight depending on desired sharpness). Uplight = softer, less artifacting. Upbeta = more aggressive detail.
- For ecommerce print or high-res needs, take the MJ upscale then run one pass in a dedicated upscaler (Topaz Gigapixel or similar) at 2x–4x to preserve edges and color accuracy.
- Always check edges and shadow continuity after upscaling; if edge halos appear, re-generate with slightly less contrast in prompt (“softer specular highlights”).
Decision criteria (pick method by need)
- Use anchor image + API automation if you need exact per-SKU consistency for 100+ shots (recommended for catalogs). Requires more setup but highest consistency.
- Use seed + exact phrasing if you can’t upload images and want simpler workflow; expect slightly more variation.
- Budget/quality tradeoff: --q 2 + external upscaler = best quality but highest cost and postwork.
Practical checklist
1) Create one perfect anchor shot (or decide exact camera phrasing). 2) Choose aspect ratio (1:1 recommended). 3) Pick a seed number and stylize (--s 30). 4) Generate small batches, pick best variant, upscale. 5) Run external upscaler if needed. 6) Batch-download and QA color/scale across all images.
Best-for / Avoid-if
- Best for: mid/large catalogs that need consistent studio-like product photos without on-set shoots. Great when you can automate via the Midjourney API or use an anchor image.
- Avoid if: you need pixel-perfect real product matching (use real photography), or if you have very complex reflections/materials that Midjourney can’t reproduce reliably.
If you want, I can generate a CSV of 100 prompt lines given a list of SKU names/colors (I can use ChatGPT to expand product briefs into consistent short prompts).
Compare Midjourney and Leonardo AI