Recommendation
Use a single reusable Leonardo “scene” (background + camera + lighting + export preset) as your template, then replace the product area with Image-to-Image / Inpainting runs driven by a CSV and the Leonardo API. That keeps angle, background and file settings consistent while letting you batch-produce hundreds of SKUs.

Decision criteria (choose before you start)
- Output quality vs cost: higher-res passes and manual touch-ups cost more time and credits.
- Technical skill: automation requires scripting (Python/node) to call Leonardo’s API and handle images; non-coders will need more manual steps.
- Time/scale: if >100 SKUs, automate; if <20, manual templating + variations may be faster.

Stepwise workflow for designers (practical)
1) Create a reference scene template
- Build one scene in Leonardo with your exact camera angle, lighting, and background. Save as a project template or export the reference image. Set final resolution and file type (PNG/TIFF with alpha if you need cutouts).
2) Prepare a clean mask area for the product
- In the scene/template, isolate the area where the product sits and export a mask or blank area (transparent/solid color). This is what you’ll preserve when inpainting.
3) Produce a product source library
- For each SKU have a high-quality front/side view or transparent PNG. If you only have flat renders, ensure consistent perspective—preferably taken from same camera angle.
4) Build prompts and negative prompts
- Create one stable prompt that specifies angle/lighting ("shot at 30° three-quarter angle, studio lighting, soft shadows"), plus a small variable section that inserts SKU attributes (color, pattern, logo placement).
5) Test and lock hyperparams
- Do a small grid (5–10 SKUs) to lock seeds, CFG/strength, and sampler so outputs are consistent. Save the successful seed/settings in the template.
6) Automate with CSV → API pipeline
- Script reads a CSV (sku, color, image_url, extra_prompt), uploads product image and mask, calls Leonardo’s API for Image-to-Image/inpainting with the locked settings, and saves outputs using your naming convention.
7) Post-processing and exports
- Use ImageMagick or a simple script to batch-resize, add bleed/margins, convert color profile, and produce final deliverables (PNG with alpha, 300dpi TIFF). Embed metadata (SKU, variant).
8) QA and sample checks
- Randomly inspect 5% of outputs. Use a visual diff or perceptual-hash to flag duplicates and check consistent shadow/scale.

Automation & templating tips
- Save the scene and parameter set as the single source of truth.
- Keep the product placement mask consistent in pixel coordinates across templates.
- Use batch naming conventions and folder structure: /raw_inputs, /ai_outputs, /final.
- Let ChatGPT help generate the variable prompt lines and a starter Python script for API calls.

Checklist (pre-launch)
- [ ] Scene template saved with resolution/export preset
- [ ] Mask area consistent and tested
- [ ] Prompt + negative prompt finalized
- [ ] Seed and strength parameters locked
- [ ] CSV format agreed and populated
- [ ] Automation script tested on 10 SKUs
- [ ] Post-process pipeline set
- [ ] QA sampling plan defined

Best-for
- Rapidly scaling catalog mockups when photography is expensive or slow.

Avoid-if
- You need pixel-perfect product photography (e.g., luxury goods where studio shots are required), or you lack the budget/skills to script and validate automation.

Final note
Start with a small pilot (10–20 SKUs) to lock camera/seed/settings, then scale via the API. If helpful, I can provide a starter CSV schema and Python pseudo-code to call Leonardo’s API.

Compare Leonardo AI and Midjourney