Uspeaks

The Uspeaks registry requires: an approved issuer signature over the asset payload, and an owner signature over the registration intent; issuer keys are explicitly registered and can be revoked; and the owner's signing intent carries a nonce to prevent replay.

Uspeaks is building voice infrastructure aimed at making control follow the voice into every machine action, with evidence passed along with the output, and a long-tail economy that keeps individuals connected to the value created by their voices.

Uspeaks precisely snapshots membership for frozen versions, ensuring buyers receive the version that was originally licensed rather than a live dataset that has since changed.

Uspeaks divides exports into three categories—analytical, foundational, and transformative—each with its own dedicated manifest, quote, license grant, and storage identifier.

Uspeaks uses an ordered roster of voice_session_ids in its ML export worker as the authoritative source for dataset membership.

Four failures remain in Uspeaks's broader two-file test suite, involving expected rollback behavior for existing export states and fallback behavior for unresolved sessions.

An analytics package containing only derived data and a transformative package containing watermarked audio differ in their permissions, risks, deliverables, and economic models, and should not overwrite each other under a single generic identifier.

Uspeaks's regression tests for canonical JSON membership, asset-audio fallback, legacy member compatibility, rolling version resolution, and frozen version snapshots have all passed.

Uspeaks's export worker resolves each session into its corresponding assets, voiceprint hash, license template, royalty data, and audio location before exporting.