What the work actually is

This is not an annotation or evaluation task — it is a one-off personal data contribution. You are asked to dig through your own camera roll and cloud backups and assemble 30 selfies and 10 short video recordings of your face, distributed across roughly the last five years. The value to the buyer is longitudinal: the same face at different ages, weights, haircuts, lighting conditions and camera generations. That is why the five-year span matters more than image beauty, and why a batch of 30 photos all taken last month will be rejected.

The hard constraint is that everything must be unedited. No Instagram or Snapchat filters, no beauty-mode smoothing, no skin retouching, no cropping that removes facial geometry. Many phones apply beautification by default on the front camera, so you may need to re-export originals from iCloud or Google Photos rather than share the versions that passed through a social app. If you can only supply filtered copies of a given photo, leave it out. Submissions are reviewed, and you should expect at least one round of feedback asking you to replace files that failed the check.

What the platform screens for

micro1's screen is short and factual. It is checking whether you genuinely hold a qualifying archive — not whether you can take good photos going forward. Expect questions about how far back your oldest usable image goes, what device it came from, whether you can access original unprocessed files, and whether you understand what you are consenting to. Because the dataset is about identity, they also confirm that every face in every submission is yours, that you are of legal age, and that you accept your likeness being used to train biometric models. If you have a sibling or a twin, saying so is an advantage — matched-relative faces are unusually valuable for this kind of dataset — but each person applies and is paid on their own account.

Logistics and honest caveats

  • Fully remote, fully asynchronous. Realistically a few hours of sorting, exporting and uploading, not ongoing work.
  • Compensation is a one-time payment on completion of the full set; the amount is not disclosed publicly on this listing and partial submissions are generally not paid. Treat any figure you see quoted elsewhere as unverified.
  • A $25 referral bonus is advertised separately from project pay.
  • Think carefully before applying. Facial data licensing is usually broad and effectively permanent — you cannot meaningfully recall images once they are in a training corpus. Read the consent terms rather than clicking through, and decline if the scope is wider than you are comfortable with.