Recommendation
Use Leonardo AI as your primary engine: generate one high-fidelity base product render (consistent lighting/perspective) and then batch-generate 100 color/pattern variants via controlled image-to-image + masked inpainting or pattern overlays. This keeps per-image costs low while preserving consistency.

Why this approach works
- Image-to-image + mask/inpainting lets you lock lighting, shadows and viewpoint and only change surface color or texture. That’s far cheaper and more consistent than re-rendering from scratch. Leonardo AI's control-to-price balance is well suited to this.

Decision criteria (pick based on your constraints)
- Fidelity vs cost: If you need pixel-perfect, brand-accurate color for commerce, studio photos are safer. If visuals are for catalog mockups or ads, AI variants usually suffice.
- Volume: For 100+ variants, automation and API batching matter. Use lower-res test runs, then upscale final winners.
- Team skill: If you have someone comfortable with masks, batch scripts, and color profiles you can maximize savings.
- Workflow stage & speed: Use AI for prototyping and marketing assets; use photography for final, regulated product images.

Practical checklist (step-by-step)
1) Capture/prepare an anchor image
- Use one clean product photo or a high-quality AI base render with neutral background, good edge definition, and shadow consistent with your brand lighting. Save a flattened and a layered/masked version.
2) Create a clean product mask
- Export an accurate mask isolating only the product surface (not cast shadow or background). This is what you’ll inpaint/replace.
3) Build a prompt template
- Example: “Replace masked area with , keep original lighting, fabric texture detail, crisp edges, consistent perspective.” Keep negative prompts to avoid background changes.
4) Prepare color/pattern assets
- For solid colors: a CSV of hex codes or names. For patterns: flattened PNGs scaled to the product UV; keep seamless tiles and alpha.
5) Run small tests (fast + cheap)
- Do 5–10 test variants at lower resolution to tune prompt strength (image-to-image strength ~0.3–0.6), guidance scale, and seed stability.
6) Batch generate using Leonardo’s UI or API
- Use the mask + prompt template, feed the list of colors/patterns, fix seed or use controlled incremental seeds to keep consistency.
7) Post-process and QA
- Apply a uniform LUT or small color correction pass to normalize any tiny lighting shifts. Check color accuracy against swatches (use delta-E tolerance if required).
8) Export, resize, and tag
- Export final sizes for web/ads, keep originals and record metadata (seed, prompt, pattern file) for reproducibility.

Best-for
- Startups that need large numbers of visually consistent mockups fast and affordably.
- Teams that can do basic masking and batch automation.

Avoid-if
- You must guarantee exact ISO color-match for regulated marketplaces or product specs.
- You don’t have anyone able to QA and correct color deltas.

Notes on budget and team size
- Smaller teams: do automated batches and manual QA on a subset. Larger teams: designate a QA lead for color accuracy. If budget is tight, lower resolution + upscaling is an effective gamble.

If you want, I can draft a sample prompt template and a CSV format for batch submission to Leonardo AI.

Compare Leonardo AI and Midjourney