Generating images

This page explains what happens between clicking Generate and the run appearing in your Reviews queue. Understanding the pipeline helps you make smarter decisions about credits and run size.

What one generation actually does

For each (product, framing, side) combination you queued, PackScene runs this pipeline:

1. Read the packshot from Shopify's CDN.
2. Run the AI generation step. PackScene calls a vision-capable model with your packshot, the framing choice, and the model gender. The result is a 1024x1024 image of a model wearing the garment.
3. Upscale the output to 2048x2048 with a high-quality image upscaler. This is where fabric texture, fine print detail, and stitch lines come back. Without this step the image looks soft.
4. Crop to 2000x2000. Center crop with no resampling, so the final image is sharp and consistent in dimensions across every product.
5. Store the result against the run. The image waits in a pending → processing → completed state machine until you approve or "don't use" it.

Each finished image consumes one generation credit from your monthly allowance. Free, Starter, Creator, and Pro tiers have different credit pools. See Plans and credits for the numbers.

What "framing" actually controls

Framing is the crop the AI generates. PackScene supports:

• Full body - head to feet. Best for dresses, jumpsuits, swimwear, anything that hangs.
• Upper body - chest to waist. Best for tops, hoodies, shirts, jackets, bras.
• Lower body - waist to feet. Best for pants, jeans, shorts, skirts, swim bottoms.

Each framing on the same product is a separate credit. You can run a single product with multiple framings in one go - PackScene generates each as its own image, all upscaled, all reviewable individually.

Front + back consistency

When you select a back packshot, PackScene generates in a specific order:

1. The front is generated first as a standalone result.
2. The completed front image is fed back in as a reference to the back generation.
3. The back is generated with the same model, same framing, same lighting.

The practical effect: front and back read as the same person from two angles, not two different people. This matters for product detail pages where customers flip between the two.

If you ever see front + back that don't match, run Rerun on that product from the Reviews page - it regenerates both sides with the same reference logic.

Batch generation

You can queue dozens of products in a single run. The Generate page is built for this: pick a collection, mark fronts (and optional backs) across many product rows, hit Generate Selected once.

The run page (Reviews → run-id) shows progress in real time. Jobs that finish appear in the review list immediately while others are still processing. You can start reviewing approved-looking ones before the run is fully done.

Practical batching advice:

• Keep collections homogeneous. A run with 30 t-shirts at upper-body framing is consistent and predictable. A run mixing dresses and shoes is not.
• Start with hero products. Generate your top sellers first to validate the look before committing credits to the long tail.
• Save a few credits for review. If a run produces a few off-brand outputs, you'll need regeneration credits to fix them. Don't drain your monthly allowance on a single batch.

What happens if a generation fails

A generation can fail for a handful of reasons: the source packshot 404s, the AI returns an error, the upscaler rejects the image. PackScene marks failed jobs with failed status in the run and never debits the credit. You can rerun the product from the Reviews page once you've fixed the input (for example, re-uploaded a clean packshot to Shopify).

Tracking credit usage

The PackScene Pricing page (Settings → Pricing in the app) shows:

• Your current plan tier.
• Generation credits remaining this month.
• Regeneration credits remaining this month.
• Total credits used this billing cycle.

Credits roll over only on the Pro tier. On Starter and Creator, unused credits expire at the end of the month. See Plans and credits.

What's next

• Reviewing images - approve, don't-use, regenerate.
• Publishing to Shopify - getting the approved images onto your products.