Short summary / recommendation
Use Leonardo AI to batch-render base product images and variations, and a small automation script (or ChatGPT prompts) to expand a SKU CSV into per-angle generation instructions. Produce one consistent studio template (lighting, camera, background), then generate per-SKU multi-angle renders with identical scene settings and export standardized PNGs + a metadata CSV for PIM import.

Step-by-step workflow
1) Prep data and references
- Create a CSV with fields: SKU, product_name, color, material, reference_image_url (optional), base_prompt_notes.
- Collect 1–2 high-quality reference photos per SKU if available (for image-to-image consistency). If none, note accurate physical descriptors (dimensions, color, texture).

2) Build a single “studio” scene template
- Fixed camera distances and lens: e.g., 50mm, distance so object occupies same % of frame.
- Lighting: neutral 3-point softbox (key top-left ~45°, fill opposite, subtle rim from behind). Or use a consistent HDRI studio map.
- Background: pure white (#ffffff) or 18% gray depending on PIM requirements. Use alpha if you need cutouts.
- Environment params: same exposure, shadows, and color profile (sRGB).

3) Prompt template + angle modifiers (use tokens)
Base prompt (replace tokens):
"Photoreal studio product photograph of {product_name}, {color}, {material}. Seamless {background} backdrop, consistent soft 3-point lighting, high detail, realistic texture, no logos or text, no people. Shot with 50mm lens, f/8, ISO100. Output: ultra-detailed commercial product shot."
Angle modifiers (append per angle):
- Front: "front view, camera centered, product upright."
- 45-degree: "45-degree front-left angle, camera slightly elevated 10°."
- Back: "rear view, camera centered."
Negative prompts: "blur, watermark, text, human, logo, low-detail, distorted proportions."

4) Batch-generation strategy
- For each CSV row, programmatically assemble three prompts (base + angle modifier). If you have reference images, use image-to-image with low-to-medium strength to preserve the reference while keeping template lighting.
- Keep seed fixed per SKU to improve consistency across angles (same seed, different angle tokens). If Leonardo AI supports batch APIs or templates, use that to submit jobs.

5) Settings & export
- Resolution: 2048–4096 px on longest edge (choose higher if print-quality needed).
- File format: PNG with alpha if you need background removal; otherwise high-quality PNG or 16-bit TIFF for archival.
- Color: sRGB, 300 DPI (PIM often ignores DPI but keep it consistent).
- Naming: SKU_angle_seed.png (e.g., 12345_front_0001.png).
- Metadata CSV columns: filename, SKU, angle, prompt_used, seed, lighting_preset, export_resolution.

Decision criteria
- Budget: use Leonardo AI subscription + API for large batches; pay-per-image can get expensive for 100 SKUs × 3 angles. If budget constrained, reduce resolution or run fewer angles initially.
- Skill level: if you have designers comfortable with retouching, prefer image-to-image + seed control. If not, generate clean renders with white background to minimize retouching.
- Team size/workflow stage: small teams should batch in smaller chunks (10–20 SKUs) to QC before full run.
- Output quality: for catalog-level photorealism invest in higher-res runs and manual QC/retouch.

Best-for / Avoid-if
- Best for: e-commerce catalogs where quick consistent visual tone is needed across many SKUs. Good when you can standardize lighting and use programmatic prompt expansion.
- Avoid if: you must match exact, highly accurate physical color/fit to mm tolerance without physical photography—AI renders may need spot checks and color-proofing.

Practical checklist before full run
- [ ] Finalize CSV with all SKU fields
- [ ] Create and lock studio template parameters
- [ ] Save one QC SKU with reference and run 3 angles, review color/scale
- [ ] Confirm seed usage and naming convention
- [ ] Export PNG/TIFF and produce metadata CSV
- [ ] Batch-run remaining SKUs in groups, review samples

If you want, I can convert your CSV into ready-to-run Leonardo prompts and a small script to assemble jobs. CTA: try Leonardo AI for the render step.

Compare Leonardo AI and Midjourney