2026-08-13

The AI action figure trend, without uploading your face to get it

In April 2025, OpenAI shipped a new GPT-4o-powered image generator inside ChatGPT, and within days social feeds filled up with a very specific image: people reimagined as glossy plastic collectible toys, boxed in a blister pack with a handful of accessories chosen to represent their personality or job. Forbes covered it as "the AI action figure trend," and the mechanics are simple — you upload a selfie, add a prompt describing the figure and its accessories, and ChatGPT hands back the box art. It's a fun, low-effort image. It's also, mechanically, a photo of your face uploaded to a third party specifically so their model can process it — and more than one privacy researcher went on record about what that specific step actually costs. Here's what they said, and a way to get the same "boxed collectible" look by cutting out a real photo of yourself instead, with nothing uploaded at any step.

What "just upload a selfie and a prompt" actually means

Jennifer King, a privacy fellow at Stanford's Institute for Human-Centered AI, told reporters covering the trend that OpenAI's default setting allows uploaded images to be used for future model training unless a user actively opts out — and her broader point wasn't about that one setting, it was about what happens after: "you just don't really know what happens after you share the data." Separately, Dave Chronister, CEO of the security firm Parameter Security, flagged a more specific angle most people don't think about: the accessories people choose to put in their own action-figure box — a guitar, a stethoscope, a particular sports jersey — are a self-authored list of "here are the three or four things I'm most interested in," which is exactly the kind of detail that makes a personalized phishing email convincing. Neither claim is about the image itself being dangerous; both are about what a face-plus-context upload to a third-party model quietly hands over beyond the picture you asked for.

The DIY version: same box-art look, nothing leaves your device

None of that is an argument against the toy-shelf look — it's an argument against getting it by uploading your photo to someone else's model. The look itself doesn't require AI generation at all: it's a photo of you, cut cleanly out of its background, sitting inside a box or blister-card template. That's exactly the two-step job NearBG is built for, minus the box art:

1. Take (or pick) a photo of yourself in whatever pose you want your "figure" posed in — even lighting and a plain backdrop behind you give the segmentation model the cleanest edge to work with.
2. Drop it into NearBG. The on-device U2Netp model (covered in this blog's first post) removes the background automatically, entirely inside your browser tab — no upload, no third-party server involved at any point.
3. Use the manual touch-up brush to erase any leftover fringe around hair or clothing edges before you download the transparent PNG.

That transparent cutout is the "figure" half of the image. The other half — the box, the blister-pack outline, the accessory icons — is a template, not something NearBG generates. Free and paid box-template packs exist specifically for this look (packaging sites like Pacdora publish free printable action-figure box templates, and Canva has its own template library), and dropping a transparent PNG into one of those in any design tool is a normal, five-minute layout step, not a specialized skill.

What NearBG does here, and what it doesn't

Worth being precise about the boundary, the same way this blog is about every other tool NearBG gets compared to: NearBG does not generate a plastic, toy-like rendering of you the way ChatGPT's image model does — it doesn't add glossy highlights, doesn't restyle your face, and doesn't invent a pose you didn't already have in the source photo. It also doesn't build the box or blister card around the cutout; that's a template you place the PNG into yourself, in whatever design tool you're already using. What it does do is the one step that's identical either way — getting a clean, alpha-transparent cutout of a real photo of you — without that photo ever being sent anywhere to get it. Worth noting for anyone reaching for Canva to do that same cutout step: Canva's own pricing page lists its Background Remover as a Canva Pro feature, not something the free tier gives unlimited access to. NearBG's automatic removal and touch-up brush are free, with no account and no daily cap, for exactly this one step.

A practical checklist

• Shoot the source photo with even, front-facing light and a plain backdrop — the cleaner the edge, the less touch-up work afterward.
• Run it through NearBG's automatic removal, then zoom in along the silhouette and use the brush on any stray background pixels the model left behind, especially around hair.
• Download the transparent PNG and drop it into a free action-figure box or blister-card template (packaging-template sites and general design tools both have them) in whatever layout tool you already use.
• If you use ChatGPT or another AI generator for the box art or plastic-toy styling itself, check that platform's own current data-training and retention settings first — that's a separate, platform-specific decision NearBG has no part in, and this post isn't a substitute for reading it.

The honest summary

The action-figure trend went viral specifically because of how easy the upload-a-selfie-and-a- prompt version is, and that same ease is what a Stanford privacy researcher and a security-firm CEO both flagged real, specific costs for on the record — a training-data default you have to opt out of, and a self-authored list of your own interests handed to a third party's model. The box-art styling itself isn't something NearBG does, and this post doesn't pretend otherwise. What it does do is the one part of the process that's identical either way — a clean, transparent cutout of a real photo of you — without your face ever being the payload of an upload.