Troubleshooting PackScene

PackScene problems almost always have one of four root causes: input image issues, account/credit issues, Shopify-side issues, or a transient AI service hiccup. This page walks the most common ones.

Generation fails for a specific product

Symptoms: the run completes, but one or more products show a failed status in Reviews.

Most common causes:

• The packshot is corrupt or 404s. Open the product in Shopify and check the image is still there. If you re-uploaded it after the run started, retry.
• The packshot is too small. Anything under ~512px on a side struggles. Replace with a higher-resolution version.
• The packshot is the wrong format. PackScene supports JPEG and PNG. SVGs, WebPs with unusual color profiles, and HEIC can fail.
• AI service hiccup. Sometimes the underlying AI service throws a transient error. Click Rerun on the affected product (uses a regeneration credit) and the issue usually clears.

Failed generations do not consume generation credits, only successful ones.

Generation is slow

PackScene queues each image and runs them in parallel up to a per-plan concurrency. If you queued 50 products at once, expect the run to take several minutes end-to-end. Larger runs are designed to be left alone - close the tab and check back. Generation continues server-side.

If a run has been "running" for more than 20 minutes and shows no progress, something is wrong - email support@upsellshark.com with the run ID.

Images don't appear on the product

Symptoms: you clicked Publish, the run flipped to accepted, but the new images aren't on the Shopify product.

Check, in order:

1. Refresh the product page in Shopify admin. The image upload finishes in the background and can take a few seconds to show up.
2. Check the position. If you set "After existing images", the AI image is at the end of the media list, not the start. Scroll all the way down.
3. Check the product is actually the one PackScene generated for. It's surprisingly easy to be looking at the wrong product. The Library page in PackScene shows which products have published AI images.
4. Check Shopify storage quota. Very rare, but if your store is at the Shopify file storage limit, uploads silently fail. The image won't appear.

If none of these explain it, contact support with the run ID.

"Out of credits" error mid-run

Symptoms: the run starts, processes some images, then stops with an out-of-credits error.

Cause: the run needed more credits than your plan allowance. PackScene debits credits as it processes, so a run that needs 50 credits but you have 30 will process the first 30 and stop.

Fix:

1. Open the Pricing page in PackScene.
2. Upgrade to a higher tier (immediate, no cycle delay).
3. The remaining images in the failed run won't auto-resume - start a fresh run for the unprocessed products, or click Rerun on them individually (uses regeneration credits, which are cheaper).

"Back image looks like a different person"

Symptoms: front and back AI images for the same product show two visibly different models.

Cause: rare, but can happen if the AI service returns an unexpected output for the back generation step.

Fix: click Rerun on the product in Reviews. The rerun regenerates the front first, then re-feeds it to the back generation, restoring consistency. Costs one regeneration credit per affected image.

"The fabric texture looks soft"

Symptoms: the model image is correct but the garment looks slightly blurry or low-detail.

Cause: usually the input packshot was lower resolution than expected, or had heavy JPEG compression that the upscaler couldn't recover.

Fix: replace the packshot in Shopify with a higher-quality original (uncropped, less compressed), then Rerun the product.

"Model identity changes between runs"

Symptoms: you reran a product or started a new run and the model looks different from previous outputs for the same product.

Cause: this is expected. Generative AI has natural variance. PackScene maintains consistency within a run (front + back use the same reference) but not across runs.

Workarounds:

• Generate front + back in one run rather than two separate runs.
• Pick one PackScene image per product and stick with it - don't keep regenerating "looking for the perfect one".
• For a uniform storefront, generate a whole collection in one run with the same gender and framing.

"The image looks wrong - wrong gender, wrong framing"

Symptoms: you picked Male, full body, but the output is Female, upper body.

Cause: extremely rare. Almost always means the wrong product was selected when generating.

Fix: open the run in Reviews. Check which product the image is attached to. If it's the wrong one, mark the image "Don't use" and Rerun. If it really is the right product, contact support with the run ID and screenshots.

Can I get a refund for failed or bad images?

Failed generations don't consume credits in the first place. For images that succeeded but you don't want to publish, mark them "Don't use" - you've still spent the credit, but you didn't waste a publish slot.

For systematic problems (whole runs producing garbage, billing charged but no credits granted), email support@upsellshark.com and we'll sort it manually.

Where the workers run

PackScene's background workers handle every generation, upscale, and publish step asynchronously. You can close the tab at any point during a run and come back later - state is server-side. If the app feels unresponsive, refreshing the page usually resolves it; the underlying work continues.

See also

• Generating images - how the pipeline works.
• Plans and credits - credit math.
• Best practices - preventing problems before they happen.