Recommendation (short)
Use Leonardo AI to create 3–5 angle templates (front, 45°, top/flat lay, lifestyle) and batch-generate 3–5 seed variations per SKU. Keep prompts templated and seeds deterministic so outputs are consistent and predictable for post-editing.

Decision criteria
- Budget: more variations and higher resolution = higher compute costs. If budget-limited, reduce variations per SKU or resolution.
- Skill level: beginners should use template prompts + reference images; advanced users can tune sampler/steps and use the API for automation.
- Workflow stage/team size: small teams—fewer templates and manual QA. Large teams—parallel batch jobs and strict naming conventions for handoff.

Step-by-step workflow
1) Prepare assets and metadata
- Collect one clean product photo or CAD render per SKU (if available) as a reference.
- Create a CSV with columns: SKU, PRODUCT_NAME, COLOR, MATERIAL, PRIMARY_PROMPT_TOKEN.

2) Design 3–5 angle templates
- Example templates: Front (90°), 45° three-quarter, Top (flat-lay), Close-up detail. Keep lighting and camera tokens consistent across templates: e.g. “studio lighting, white seamless background, soft shadow, 85mm lens, f/4, ultra-realistic”.
- Create one reference image per angle (a neutral product example) to use for img2img conditioning if you want deterministic geometry.

3) Create a prompt template
- Example: “{PRODUCT_NAME} in {COLOR} on a white seamless background, studio lighting, realistic texture, 45° angle, subtle contact shadow, high detail, 85mm lens; no text, no extra props.”
- Use negative prompts like: “no logos, no people, no blur.”

4) Seed and variation strategy
- Assign a base seed per SKU (e.g. hash of SKU) for determinism. Generate N variations by adding offsets (+1, +2…). Keep guidance/CFG and steps fixed across the project.
- Variation count: 3–5 per angle (so 9–25 images/SKU depending on template count).

5) Batch generation and automation
- Use Leonardo’s batch/project interface or API to loop the CSV: for each SKU and angle, inject prompt tokens, attach reference image (if using), set seed, request X variations.
- Run in batches (10–20 SKUs per batch) for easier QA and re-run of failed SKUs.

Export settings (recommended for post-editing)
- Resolution: 3000x3000 px (square) or 3000 px longest side. If file size is a concern, 2048x2048 is acceptable.
- Format: PNG with alpha (transparent background) if you want instant compositing; otherwise high-quality PNG on white.
- Steps: 20–40; Guidance/CFG: 6–9.
- Variations: keep small changes to preserve geometry.
- Naming: SKU_ANGLE_SEED_VARIANT.png (e.g. SKU123_45deg_10021_v01.png)

Checklist before large run
- [ ] CSV populated with all SKU tokens
- [ ] 3–5 angle templates finalized and tested (QA 2 SKUs each)
- [ ] Reference images prepared for each angle
- [ ] Seed scheme and naming convention decided
- [ ] Export settings validated for post-editing (transparent PNG, resolution)
- [ ] Batch size & compute budget confirmed

Best-for / Avoid-if
- Best for: ecommerce teams needing many consistent hero shots or detail angles ready for retouch/compositing.
- Avoid if: you need perfect pixel-for-pixel geometry for CAD-to-photo matches (use actual product photography or CAD renders) or if you lack budget for multiple high-res generations.

Practical tip
Use ChatGPT to generate prompt permutations and to create the CSV rows quickly (one tool to help automate prompt expansion). Then run the actual image generation in Leonardo AI and keep a strict QA pass (visual check of geometry and shadows) before sending to retouching.

If you want, I can draft the exact prompt templates and a sample CSV row for 3 angles you can paste into Leonardo’s batch tool.

Compare Leonardo AI and Midjourney