Every major tech company — Meta, Google, TikTok, YouTube — uses facial demographic analysis to profile you, sort your feed, and target you with ads. They just don't show you the data. Volkus flips the script. Your ethnicity, age, gender, and skin tone travel publicly with every post. Anonymous voice. Transparent identity.
On Volkus, posts look like this:
No account. No email. No real name. Just a 30-second face scan, a demographic label, and a seat at the table. Your identity is preserved by math, not by surveillance.
You agree to a 30-second webcam session. The analysis runs entirely in your browser. No video is uploaded anywhere.
Our Folkus AI model assigns you an ethnicity, age bracket, gender, and skin tone (Fitzpatrick scale). This becomes your public tag.
Your device fingerprint (fonts, hardware, location) creates a unique token. You're remembered without an account.
Post anonymously on any board. Every vote — up or down — carries the voter's demographic. The data is visible to all.
Filter posts by any demographic combination. Only Zoomer takes? Only South Asian threads? Only lighter or darker Fitzpatrick zones? Done.
The age model must return ≥ 18 years to proceed. Under-18 detections are blocked at the scan gate.
↑ Hover any vote count to see a demographic breakdown of who voted — age, ethnicity, gender, country.
The demographic analyser was built from the ground up as part of the Smart Glasses Lab project. It uses a suite of custom-trained MobileNetV2 ONNX models called Folkus, running entirely client-side in your browser via WebAssembly. No face data ever leaves your device.
Detection is powered by MediaPipe FaceMesh — the same technology used in Google Meet, Instagram filters, and Snapchat lenses — which extracts 468 3D facial landmarks in real time. The Folkus models then classify each face crop independently for ethnicity, age, gender, and skin tone.
Top-1 accuracy across 7 classes. Folkus-Race (MobileNetV2), trained on FairFace 86k images.
Binary classification. Folkus-Gender. 92.1% overall (Male 91.8%, Female 92.5%).
Continuous age regression. Folkus-AgeFused, conditioned on detected ethnicity & gender.
Per-face inference in-browser. 30-second rolling average for stable classification.
ITA-based skin colour classification via forehead pixel sampling. Outputs Fitzpatrick zone (I–VI), sampled RGB, and ITA angle. Separate from and complementary to ethnicity.
| Label | Skin Tone | Notes |
|---|---|---|
| White | Ivory–Fair | European, Anglo, Slavic |
| Black | Deep Brown | African, Caribbean, African-American |
| East Asian | Peach–Warm | Chinese, Japanese, Korean |
| South Asian | Golden Brown | Indian, Pakistani, Bangladeshi |
| Latino / Hispanic | Warm Beige | Sociolinguistic label — covers all phenotypes |
| Middle Eastern | Medium–Deep | Arab, Iranian, Turkish, Levantine |
| Mixed / Other | Variable | When top confidence is below threshold — no forced label |
| Label | Age Range |
|---|---|
| Zoomer | Under 30 (Gen Z) |
| Millennial | 30–44 |
| Gen X | 45–60 |
| Boomer | 61+ |
Meta ran facial recognition on every face in every photo you uploaded — confirmed by $650 million and $1.4 billion in legal settlements for doing it without consent.[1][2] Google openly documents that it infers your age and gender from search and browsing behaviour to segment you for advertisers — you can see your own inferred profile in your Google account.[3] TikTok explicitly updated its privacy policy to claim the right to collect "faceprints and voiceprints", and leaked internal guidelines confirmed algorithmic filtering by inferred user appearance.[4][5] The FTC's 2024 surveillance report named all three as conducting demographic profiling beyond what users knowingly disclosed.[6] This technology is already everywhere. It just doesn't show you the output.
Volkus doesn't hide the analysis. We surface it — make it community knowledge. The data becomes visible to everyone, including yourself. That's the only thing we've changed. We are a mirror, not a machine.
When you know your demographic is visible, something interesting happens to what you say — and how others receive it. That's the experiment.
Volkus is a petri dish. We are watching what happens when you take the invisible demographic metadata already present in every online conversation and make it visible.
Every upvote and downvote carries the voter's age, ethnicity, and gender. See exactly which demographics are amplifying or suppressing which voices.
Want only Zoomer takes? Only Black & East Asian threads? Only posts from Millennial Women? Set it. See the difference in discourse.
Your GeoIP country displays as a flag on every post — like 4chan's /pol/ — so geographic origin shapes the conversation visibly.
Disagree with your assigned label? You can set your own. To you, your posts show your choice. To everyone else? "Black 31y M → Thinks they are → White 22y F."
Aggregate patterns — who votes for whom, posting habits, cross-demographic engagement — published openly for all to analyse.
An AI analyses trending topics and seeds discussion threads to get conversations started across boards.
Volkus exists to observe how demographic identity shapes speech, reception, and community — when that identity is visible rather than hidden. The experiment only works if people post honestly. These rules are short. Read them.
Volkus is built on radical anonymity combined with radical demographic transparency. These are not in conflict. The analysis that labels you also protects you — because the label is all we retain.
This platform is operated from Australia and governed by Australian law, including the Privacy Act 1988 (Cth) and the Australian Privacy Principles (APPs). Users outside Australia access the service subject to these terms. For privacy enquiries: [contact to be added]. Last updated: 2026-04-22
As of the date below, Volkus has not received any national security letters, FISA court orders, gag orders, or any other compulsory legal demand requiring non-disclosure. We have not been compelled to provide backdoor access to any system or user data. We have not been subject to any surveillance or interception orders we are prohibited from disclosing. No user data has been handed to any government, law enforcement, or intelligence agency.
If this notice disappears or the date below stops being updated monthly, treat that as the signal that the above statements are no longer true. This is a transparency mechanism — not a legal guarantee.
Last updated: 2026-04-22
The scan takes 30 seconds. Runs entirely in your browser. Over 18 only. You don't need an account. You just need a face.
Get Analysed — Start the Scan →By proceeding you confirm you are 18+ and consent to the 30-second demographic analysis.