Aurora Canvas: what it is, what it costs, and how to start

Aurora Canvas is a browser image generator built around one workflow: you already have a photo of a real object, and you need it to look like it was shot somewhere else. That is a narrower promise than “generate anything”, and it is why the tool is useful — it refuses to invent the product, and spends its capacity on light instead.

Most image generators ask you to describe the whole scene. This one asks you to describe the difference between your photo and the photo you want. Once that clicks, the prompt gets shorter, not longer.

Who it fits

If you sell physical products and shoot them on a phone, this replaces a lightbox for most listing images. If you need photorealistic hands, faces or text inside the image, look elsewhere — the reference constraint that makes products stable makes people stiff.

For heavy local work, pair it with an open model you can run yourself: our Nimbus 7B setup guide covers the VRAM numbers and the one command that starts it.

What the free tier really covers

Sixty credits a month at one credit per image is sixty images. In practice that is one finished project: four rounds of four variants, plus retries for the two prompts that came out wrong. It is enough to decide whether the tool fits your work, which is the point of a free tier.

Two limits matter more than the credit count. Free generations may be shown in the public gallery, and batch size is capped at four. Neither is a blocker for evaluation; both are reasons to upgrade before you use it on client work.

Where it falls short

• Long prompts degrade fast. Past roughly forty words the model starts ignoring clauses, so say one thing per sentence.
• Text on packaging is unreliable. If the label is legible in the reference it usually survives; if you ask for new text, it will be mush.
• Shadows are the giveaway. Generated shadows are cleaner than real ones, so add a little grain after export.

Bottom line

For product and listing photography, this is the shortest path from a phone photo to something you can publish, and the free tier is honest about its limits. For illustration, it is the wrong tool — the reference constraint fights you the whole way.

If you want the full workflow, the product photo tutorial takes the same tool from a raw phone image to a finished listing asset.

What it actually does

• Reference-guided generation

  Upload one product photo and the model keeps the shape, label layout and material while you change the scene around it.

  

• Scene presets

  Twelve lighting setups (studio, window light, hard sun, neon) that map to prompt tokens instead of being separate models.

• Batch variants

  Generate four variations of the same prompt in one request, which is the fastest way to find a usable seed.

• Layer export

  Export subject, shadow and background as separate PNGs at 2x, so the result can be finished in a normal editor.

How to get your first result

1. Create an account and check your credits

   Sign up with email. The account page shows 60 credits, and each 1024px image costs one credit. Nothing is charged until you add a card.

2. Upload a reference photo

   Drag a photo of the object onto the canvas. Crop tightly around the product; loose crops make the model invent edges.

3. Write the scene, not the object

   The reference already carries the object, so the prompt should describe only light, surface and framing. Copy the template below.

4. Generate four variants, then export

   Pick the variant with the cleanest shadow, then export layers at 2x for retouching.

Free tier and pricing

Plan	Price	What you get	Best for
Free	$0	60 credits per month	Finishing one small project before committing
Creator	$12 per month	600 credits, private generations	Weekly product listings
Studio	$39 per month	3,000 credits, shared brand presets	Small teams shipping a catalogue

Prices checked on 2026-10-23. Plans change without notice — confirm on the official pricing page before you buy.