Batch Product Images for Free with Google AI Studio and Nano Banana
What this actually saves you
A founder running a kitchenware brand told me her product-photo line item was eating four figures a month and asked whether AI could dent it. Before hiring another retoucher, the first stop should be Google AI Studio. It is Google’s web playground for trying out models at aistudio.google.com, you sign in with a Google account, and no credit card is required to start.
The image model inside it is Gemini 2.5 Flash Image, nicknamed Nano Banana. The point that matters for ops teams is the free daily allowance, which is sizable. As of writing, published numbers vary a lot, anywhere from fifty to a thousand images a day depending on who you ask, so call it typically several hundred per day and treat it as variable by model version, region, and account rather than a fixed figure.
For a catalog with a few hundred SKUs that each need a white-background hero plus two or three lifestyle shots, that volume covers most day-to-day merchandising. You do not need to buy anything up front. Burn the free quota first, then decide. The rest of this walks through it in working order.
Getting in and choosing the right model
Open aistudio.google.com and sign in at the top right. You land in a chat-style interface: history on the left, a prompt box in the middle, and a panel on the right for the model and parameters.
The step people miss is selecting the right model. In the model dropdown on the right, pick Gemini 2.5 Flash Image, that is what puts you on Nano Banana. Leave it on a text model and you will get no image at all. Nano Banana 2 landed around February 2026 and is a clear step up from the older version. If you see a Nano Banana Pro option, that is the higher-fidelity variant with better text rendering, but in the consumer app the free allowance is typically only a couple of images per day, so save it for the shots that truly need it.
Before generating, set the aspect ratio: 1:1 for a marketplace hero, a vertical frame for a long detail page. Upload your product photo, write the prompt, hit enter, and the image comes back in a few seconds. Right-click to save it or use the download button.
Prompting so the product shots hold up
I write prompts in five ordered parts: subject, background, lighting, camera, and style. For a white-background insulated bottle I would write: a matte black stainless steel insulated bottle, pure white shadowless background, soft even top light, front-on view with a slight downward angle, ecommerce hero style, product filling about seventy percent of the frame. That structure keeps the model from improvising.
For a clean cutout, lock the background down explicitly with pure white shadowless background, otherwise it tends to add a faint gray gradient or a drop shadow that you then have to mask out. For lifestyle shots, describe the scene concretely. Do not just say on a table, say on a light oak desk next to an open book and a cup of coffee with morning light from a window on the left. Specificity is what makes it repeatable.
For try-on and on-model work, Nano Banana holds product consistency better than the previous generation. Feed it a flat-lay of the garment and ask for a model wearing it, specifying body type, pose, and setting. This is still the hard case, though. Run ten and three or four may warp a collar or a pattern, so generate several and pick. As for text on the image, such as a promo overlay, the Flash Image tier is honestly shaky. If you need accurate type, that is a job for Pro.
How to actually batch it
The web interface is click-by-click, so it is not automation, but a few habits make it much faster. First, templatize the prompt. Dial in one prompt per product category, then change only the subject line when you swap images, reusing the background, lighting, camera, and style. Output stays consistent and you stop rethinking it each time.
Second, split complex images into steps. Do not expect one sentence to produce a perfect composite. Have it cut the product to a clean white background first, then feed that clean image back in as the input and have it composite into the target scene. Two steps, each controllable, far fewer throwaways.
If you genuinely need scale, say a few hundred SKUs running unattended, drop the web interface and use the Gemini API. The API has a free tier too, on the order of several hundred images a day as of writing, callable from a script. Write a small script that reads a folder of product photos, loops through the API with your prompt template, and writes results to matching folders. Leave it running overnight. This needs a bit of code, so have an engineer write it or have an AI assistant draft the script for you.
One privacy note: on the free tier, your inputs, meaning the images and prompts you submit, may be used to improve Google’s products under their terms. If your imagery involves unreleased products or exclusive designs, keep it off the free tier. The paid tier typically excludes training use, which is the safer route for brand-sensitive work.
When the free quota runs out, and when to upgrade
The free quota resets at midnight Pacific. If you exhaust it in the morning, waiting a bit usually gets you topped back up. For teams with modest daily volume, the free tier plus a little off-peak timing means many sellers never spend a cent.
This table sorts tasks by the approach that fits:
| Task | Recommended approach | Rough cost |
|---|---|---|
| White-background and simple lifestyle shots | AI Studio web free tier, Nano Banana | Zero |
| Try-on and multi-element composites | Free tier, generate several and pick | Zero, costs time |
| Accurate promo text on the image | Nano Banana Pro | Free a couple per day, paid beyond that |
| A few hundred SKUs, unattended | Gemini API, consider the paid tier | Usage-based |
| Unreleased products, exclusive designs | Gemini API paid tier (excludes training) | Usage-based |
So when do you pay? Two cases. One, you consistently need Pro-level fidelity or accurate text and a couple of free images a day will not cut it. Two, you need steady volume and keep hitting the free ceiling even with off-peak timing. For either, move to the paid Gemini API tier, billed per token, which for modest volume runs a fraction of what outsourcing costs. Until you hit one of those triggers, just ride the free quota and do not pre-pay for demand you do not have yet.
Related Articles
Free Product Videos for TikTok and Reels with CapCut AI
A practical workflow for producing short-form product videos at zero cost using CapCut free AI features: AI script, text-to-speech voiceover, auto-captions, templates, and short AI b-roll. Includes a free-versus-paid feature split and platform-specific tips.
Amazon DSP Creative Agent: One Product Photo to a Full-Funnel Campaign
In 2026 Amazon is expanding the Creative Agent inside its Generative AI Creative Studio to more marketplaces. Sellers can turn a single product photo into display, video, audio, and Streaming TV ads. Here is what it does, how the workflow runs, and what to check before launch.
A Zero-Cost Conversion Stack: Microsoft Clarity Heatmaps Plus GA4
GA4 tells you a product page has a high exit rate, but never why visitors leave. Microsoft Clarity is free forever with no traffic cap, and it fills the heatmap and session-replay gap. Pair the two and you get a $0 behavior stack that covers nearly everything a small store needs to find conversion leaks.