How do I generate AI product photos in bulk for many SKUs at once?

Quick Answer

Nightjar does not have a CSV importer or a "select 200 SKUs and click go" button. The bulk model is reuse: save the photography style, composition, model, background, and output settings as a Recipe, then swap the product image and apply that Recipe per SKU. Two product photos from the same Recipe come out of the same shoot system instead of drifting per prompt.

Why bulk in Nightjar means reuse, not a queue

Some catalog tools lean on batch processing: upload a CSV, point at a folder, watch the queue run. Nightjar takes a different shape. Each run of the Create form, called a Generation, accepts one to five product images at a time, so a thousand SKUs is a thousand Generations.

What removes the rebriefing cost is a Recipe: a Team-owned saved Create-form setup that captures the photographic look (camera, lighting, mood), the composition (framing, angle, pose, placement), the optional model, the background, any extra written instructions, and the output settings (aspect ratio, resolution, format, image count). You build it once on your first SKU. Every SKU after that is: swap the product, apply the Recipe, generate.

The per-SKU loop

1. Open Create and pick the Product Listing Image Workflow, the path Nightjar uses for ecommerce-ready primary product imagery.
2. Upload the first SKU's product image (one to five product images per Generation).
3. Pick the visual ingredients: a Photography Style for camera, lighting, and mood; a Composition for framing, angle, and product placement; a Background; and an optional Fashion Model (a reusable AI person who appears with the product). Add any Custom Directions, Nightjar's name for the free-text refinements layered on top of the ingredients.
4. Set aspect ratio, resolution (1K, 2K, or 4K), output format (JPEG, PNG, or WebP), and image count (one to six candidates per Generation on paid plans; the default is two).
5. Click "Save as recipe" and give it a name (up to 60 characters; a Team can keep up to 100 active Recipes).
6. For the next SKU, swap the product image and apply the Recipe. Only the product changes.
7. Generated images land in the Team's Library (the shared collection of uploaded and generated Assets), where they can be downloaded, edited, upscaled, or reused as references.

The point of the Recipe is that nothing visual has to be retyped or rejudged for the second SKU. The look is locked.

Levers that turn per-SKU into a fast loop

• Up to six candidates per Generation. One product image, one click, six options to pick from.
• Recipes are Team-owned. One person can build the visual system. The rest of the Team applies it without rebuilding the brief.
• Shared Library. Every Team member sees the same product uploads, generated Assets, ingredients, and Recipes.
• Edit Shortcuts for variants. Edit Shortcuts are pre-filled prompts in the Edit tab for common edits. For colorways, format swaps, or reframes on a hero image, the Recolor, Change Format, or Reframe shortcut works on a single source instead of re-running Create per variant. The Edit board accepts up to eight reference images.
• Photoshoot for gallery expansion. Photoshoot is a separate Workflow that takes one source Asset and returns four cohesive variants that feel like one shoot, useful for filling a PDP gallery without rebuilding the brief.
• Upscale for legacy refresh. Upscale is the Workflow that brings an existing image to a 2K or 4K long edge while preserving product content. It skips work and Credits when the source already meets the target.

What Nightjar will not do today

It is worth being direct about the boundaries.

• There is no CSV import that takes a list of products and runs Generations against each.
• There is no "select 200 products and click go" multi-SKU queue.
• Generated images are not synced directly into Shopify product media. Download and upload separately.

The honest framing: Nightjar scales by reuse, not by parallel processing. If the bottleneck is keeping a 500-SKU catalog inside one visual system, a Recipe handles that better than a batch job. If the bottleneck is pure throughput at any aesthetic, a CSV-driven tool may suit the workflow more.

Credits at catalog scale

Credits are the Team-owned balance Nightjar deducts from on each Generation. Each image candidate at 1K or 2K costs one Credit. Each candidate at 4K costs two Credits. Photoshoot is a flat two Credits (four at 4K) for four related images. Plan around the candidate count: six candidates per Generation across 200 SKUs is 1,200 Credits at 2K, 2,400 at 4K. Drop the candidate count to two when you trust the Recipe, and the same catalog runs on a third of that.