Last updated: July 30, 2026
Voice Exchange — solo tasks, reading assessment & live conversations
These Voice Contributor Rules apply to Human Archive’s Voice Exchange feature. They are in addition to the Contributor Terms, the Privacy Policy, and the Contributor Recording Rules. By taking part in Voice Exchange you agree to these Rules.
The conduct rules for voice recording — where you may record, who may be audible, what you must not say aloud, and what is prohibited — are set out in Section 5 of the Contributor Recording Rules, together with Sections 6, 7 and 8 of that document. This page covers the terms specific to voice contributions.
1. What Voice Exchange records
Voice Exchange has three sub-types, and all three record audio only — no video is captured:
Your voice is recorded, uploaded, and stored. Alongside the recording we keep data derived from it by our automated review — a transcript, the detected language, a speaker-count value, audio-quality scores, and a probability score for whether the recording is likely AI-generated. We do not extract voiceprints or speaker embeddings from your recordings, and we do not use your voice to uniquely identify you.
2. Consent
Before your first Voice Exchange task we show you a one-time voice-consent screen that you must accept. Before every two-party conversation we show both participants a per-call consent screen, and no recording begins unless both have accepted. You can withdraw your consent to future Voice Exchange participation at any time in Settings → Account.
3. All-party recording
A two-party conversation records both voices. Take part only if you consent to being recorded, and do not share sensitive personal information — financial data, health data, government IDs, or other people’s personal information — during a conversation. What is said aloud is recorded, stored, and included in datasets we license to third parties.
4. Eligibility
You must be at least 18 years old. Voice Exchange is not available to residents of Illinois, Texas, Washington, or Oregon, and may be unavailable in other regions.
5. License you grant
You keep authorship of your voice recordings. You grant Human Archive a worldwide, perpetual, irrevocable, non-exclusive, royalty-free, fully paid-up, sublicensable license to host, store, transcribe, adapt, create derivative works from, and use your voice recordings and data derived from them (including transcripts, detected language, and audio-quality and related scores) to develop, train, evaluate, and improve machine-learning models, and to include them in datasets Human Archive licenses to its customers for those purposes. Voice datasets are licensed separately from video datasets.
6. Service providers
Live conversation audio is carried and recorded by LiveKit; recordings are stored and analyzed with our cloud infrastructure provider (currently Amazon Web Services); the reading-assessment clip is scored by a third-party AI service (currently Google’s Gemini API). Live conversation audio is not sent to the Gemini API. See the Privacy Policy for the current list of service providers and how each handles voice data.
7. Review, approval, and payment
Every voice submission goes through the quality-assurance review described in Section 11 of the Contributor Recording Rules. Where a payout applies, it is shown in the app. Contributors who join through a distributor network are compensated by that network, not in-app.
Human Archive may reverse or revoke an approval that was issued due to technical error, fraud, duplicate submissions, or material quality or policy issues discovered after initial review. Human Archive will notify the contributor of any such reversal. No payout is earned until a submission has received final approval under our review process.
8. Retention and deletion
Human Archive retains voice recordings and derived data for as long as needed for the purposes described in the Privacy Policy and destroys them in line with the retention practices set out there and with applicable law. You can withdraw consent and request deletion at any time in Settings → Account. Approved recordings already incorporated into datasets are de-identified — unlinked from your identity — and kept, consistent with the Contributor Terms and the applicable customer license.
9. Changes
We may update these Rules. Material changes are re-consented in the app (consent is versioned).
Questions: team@humanarchive.ai.