How long does it take to generate one AI product photo?

Quick Answer

A single AI product photo typically returns in well under a minute. The more useful number is the per-batch time: in Nightjar, one Generation (a single user-initiated request) can ask for up to six candidate images at once, and they render in parallel, so a six-candidate request finishes in roughly the same wall-clock time as a one-candidate request. Several Generations can also be queued at the same moment, which is how brands turn one saved setup into many SKUs in minutes rather than hours.

Why "per image" is the wrong unit

The literal answer for one AI product photo is short, usually under a minute. The more useful answer is that the system bills time per Generation, not per image inside the Generation.

A Generation is Nightjar's term for one user-initiated request that can produce one image or several. Inside a Generation, the candidate images dispatch concurrently rather than one after the other. The slowest candidate gates the request; the rest finish around the same moment. So a six-candidate request does not take six times longer than a one-candidate request. It takes roughly the same wall-clock time, plus a small variance from the slowest call.

That distinction matters when you are evaluating throughput. A tool that returns one image in ten seconds is not faster than a tool that returns six images in twenty-five seconds, once you account for the candidates you would have asked for anyway.

What changes the wait

The wall-clock time for a single Generation depends on a few inputs:

• Workflow. Different paths run different pipelines. A Product Listing Image Generation (Nightjar's Workflow for ecommerce-ready primary product imagery) runs a different pipeline than a Photoshoot (its Workflow that expands one input into four cohesive variants), an Edit Images Generation, or an Upscale.
• Image Type within Product Listing Image. A clean listing-oriented Generation usually returns faster than a lifestyle-oriented one, where the model has more scene to compose.
• Resolution. A 4K Generation does more work than a 1K or 2K Generation and tends to take longer.
• Upscale. Running Upscale (Nightjar's Workflow that brings an existing Asset to a 2K or 4K long edge while preserving the product) is a separate Generation on top of the original; budget for it as additional time, not as part of the first Generation.

How many images per Generation

The "1 to 6" range is specific to the Product Listing Image Workflow on the Nightjar Create tab. Other Workflows have different shapes:

If you want six different lifestyle takes on a new SKU, that is one Product Listing Image Generation. If you want to expand a hero shot into a small editorial set, that is one Photoshoot Generation, which produces four related images.

Running many products in parallel

Per-image latency is only half the story. Nightjar does not serialize Generations against each other. After you submit a Generation, the form is free immediately, so you can configure and submit the next product right away. Generations are scheduled and execute in parallel on the backend.

That is the practical mechanism behind catalog-scale work. Nightjar has a feature called Recipes: a saved Create-form setup that captures the Photography Style (camera, lighting, mood), the Composition (framing, angle, pose, placement), the Fashion Model, the background, and the output settings. A Recipe can be applied to twenty product Assets in a few minutes of clicking, and the resulting twenty Generations progress in parallel rather than queueing.

For a typical product drop with a handful of SKUs and four to six candidates per SKU, the bottleneck is reviewing and selecting outputs in the Library, not waiting for them to render.