# Seedream product photos for e-commerce: a catalog without a studio
_Disclosure: This post contains affiliate links. If you buy through them, INS.ae may earn a commission at no extra cost to you. We only recommend tools we'd use on client work._
It's Thursday night and you're on your kitchen table with a white bedsheet, a ring light you bought because a YouTube video told you to, and your phone propped against a spice jar. Forty listings need photos this week. The lampshade keeps catching your shadow, the white balance shifts every time you move the box, and the 12 SKUs you shot last month have a warmer background than the ones before them. Noon and Amazon won't reject your photos for this. Your conversion rate will.
Here's the outcome worth chasing: a clean, consistent, conversion-ready photo set for every listing, made the same day you decide to make it, without booking a studio or hiring a photographer. That's the job AI product photos for e-commerce can now actually do, and Seedream is the model I'd hand it.
What Seedream actually is
Seedream 5.0 Pro is ByteDance's image model, the one positioned for e-commerce and visual intelligence work. You don't use it through some developer console. It lives inside BytePlus Lumina, their all-in-one AI creative platform: images, video, apps, a canvas builder, all drawing from one credit balance. The image side has three Seedream versions (5.0 Pro, 5.0 Lite, 4.5), and the same account also gives you Seedance for video, audio, and a chat assistant, which matters later when your listings need motion.
Think of it as a design tool that takes instructions, not a slot machine that takes prompts. Generalist AI image generators are famous for beautiful output that's subtly wrong. A perfume bottle with five letters misspelled in the brand name, a sneaker whose sole changed shape between shots. Seedream 5.0 Pro is the version pointed at the boring, hard part of that problem: keeping your product exactly itself.
The five things it does for a catalog
Packaging fidelity. You upload the phone photo of your box, pouch, or bottle, and the model reads the label, logo, and typography as assets to preserve rather than texture to remix. This is where a generalist generator falls apart on commercial work, and it's what the Pro model is built around.
Brand VI consistency. Same background temperature, same shadow behavior, same corner radius on your lifestyle frames, across all 300 SKUs instead of the forty you had patience for. Consistency is what makes a store page look like a brand instead of a bazaar stall, and it's the thing a table-and-bedsheet shoot can never quite hold.
The multi-angle product imagery app. Front, side, three-quarter, top-down, from one source photo, as a packaged app instead of a prompt ritual. Amazon wants the alternate-angle slots filled. Noon's product pages reward it too. Filling six angle slots by hand for every SKU is how catalogs age. This app is the single most catalog-shaped feature in the platform.
The e-commerce banner creation app. The same product shots turn into category banners and promo creatives, which means you're not rebuilding a campaign visual from scratch every Ramadan and White Friday. If you'd rather someone else run that pipeline end to end, that's literally what we do at INS.ae.
Batch variations for listings. Change the background, swap the colorway, localize the packaging panel, and the composition stays put. Variant-heavy catalogs (a mug in six colors, a thobe in four fabrics) are the case where batch work pays for itself fastest, because the first generation becomes a template instead of a lottery ticket.
There's also a virtual try-on app for clothing and footwear, worth a look if apparel is your category. It gets its own deep piece in this campaign, so I won't cram it in here.
What the credits cost
Lumina runs on credits, one balance for everything in the platform. What's verified:
- A free tier with signup credits, and entry-level output carries watermarks.
- A $0.99 Starter Pass: three days of premium access, one-time purchase. This is the cheapest serious look at the model in the category, and it's enough to test on ten of your own SKUs before committing to anything.
- Subscription tiers (Basic, Standard, Advanced, Ultra) billed monthly or yearly. Subscription credits reset every month, so unused monthly credits are gone. Top-up credits are different: they stay valid for 2 years, which is the right way to cover a seasonal spike.
- Yearly billing takes 50% off the monthly price.
- Watermark removal and commercial use rights come with paid plans. On the free tier you're producing mockups, not listing images.
One thing I won't do: print a dollar figure per tier. Lumina's live tier prices sit behind a pricing app that varies by region and account, and the numbers floating around third-party listings aren't reliable enough for a recommendation. Open ai.byteplus.com/lumina/en/pricing logged into your own account and read what it actually shows you. Budget advice that survives contact with reality beats a screenshot from someone else's country.
My honest read on the model: credits resetting monthly is the gotcha. If your catalog work comes in waves (pre-season crunch, then quiet), pair a low subscription with top-ups rather than buying a heavy plan you let sit idle.
How it stacks against your other options
| Option | Cost shape | Turnaround | Consistency across 300 SKUs | E-commerce workflow |
|---|---|---|---|---|
| Product studio / photographer | Per-shoot or per-SKU quotes, money leaves before images arrive | Book, shoot, edit, wait | Excellent, that's what you're paying for | Deliverables sized for your marketplace, usually |
| Phone camera + DIY editing | Your evenings; near-zero cash | Slow, and you're the bottleneck | Weak, drifts between sessions | None, every resize is manual |
| Generalist AI image tool (e.g. Midjourney) | Subscription per seat, no usage-based tier for catalog work | Fast | Good look, shaky product accuracy on packaging and logos | No e-commerce apps, no angle or banner workflow |
| Seedream 5.0 Pro in Lumina | $0.99 trial, then credit-based subscription | Same-day batches | Built for it (brand VI consistency) | Multi-angle, banner, try-on, batch variation apps |
The studio row deserves respect, because a good studio wins on the criterion that matters most: a human controls the light, and the photo of your product is your product. For a hero image on your single bestseller, book the studio. What a studio can't survive is the math of 50 to 500 SKUs, angle slots on each, banners per campaign, refreshed seasonally. Per-shoot pricing was built for catalogs of dozens, not hundreds.
Midjourney is the fair apples-to-apples question, since people reach for it first. Midjourney makes stunning images and offers no e-commerce workflow at all. No multi-angle app, no banner app, no packaging-fidelity story. It's subscription per seat, priced in the low double digits to low triple digits per month on their public tiers, so check their pricing page before you quote the number anywhere. For a catalog job, the tool that ships with the catalog workflow wins, even if the other tool wins the art contest.
Where this loses, plainly
Credit math per task is opaque. A generation's cost moves with model, resolution, and mode, so two SKUs can bill differently. The Starter Pass exists precisely so you can measure your own per-image cost before subscribing, and I'd take that offer seriously.
Content moderation is real. Filters apply to generated imagery, which mostly won't touch bottles and boxes, but model-heavy lifestyle content can hit restrictions, and if faces matter to your listings you should test early, not mid-campaign.
It's cloud-only. No local generation, no self-hosting, nothing you can run on a plane to Riyadh. And the free tier's watermarks mean your first drafts aren't listable images; commercial rights arrive with a paid plan.
What the first week looks like
Start at Lumina with the $0.99 pass. Pick ten SKUs, the ugly middle of your catalog, not your bestsellers. Shoot one honest phone photo of each, any background, and upload it to the multi-angle app. Generate the angle set, then run the banner app on your three campaign-ready products. You'll cull. Every catalog does on the first pass, maybe a third of output gets deleted. What you're measuring is two things: does it hold your packaging text and logo, and what fraction of generations are usable per credit spent. If both answers work at day three, your full catalog becomes an afternoon job per batch instead of a month of evenings, and you can point the leftover time at automating the rest of your content pipeline.
FAQ
Can I use AI product photos on Amazon and Noon?
Marketplace rules come down to accuracy: the image must represent the product a customer receives. AI-generated scenes that keep your actual packaging and product shape intact, like Seedream's workflow is designed to do, are the safe end of that line. Anything that invents contents, quantities, or accessories you don't ship is a return-rate problem regardless of which tool made the pixel.
How much does it cost to generate product photos with Seedream?
Test access is a one-time $0.99 three-day Starter Pass. After that it's credit-based subscription tiers (Basic through Ultra) with monthly credits, top-up credits valid for 2 years, and 50% off yearly billing. Exact per-tier prices are gated behind Lumina's regional pricing page, so read them at ai.byteplus.com/lumina/en/pricing while logged into your account.
Is there a free tier for Lumina?
Yes. Signup grants free credits so you can generate before paying, but free-tier output carries watermarks and commercial use rights arrive with paid plans. Treat the free tier as a demo and the $0.99 pass as your real evaluation window.
Does it work for clothing as well as packaged products?
Lumina ships dedicated virtual try-on apps for clothing and footwear, plus multi-product try-on, so a garment on a model doesn't require a photoshoot. For apparel catalogs this is the feature to test first, and watch for the full try-on piece later in this campaign.
