Make the AI visible
AI-assisted tools are labelled and their output is presented as a proposal the user can inspect, edit or replace.
Responsible AI & trust
Useful AI requires clear boundaries.
Voxify combines generative assistance with an editable production workflow. We document what the system does, keep creative decisions reviewable and mark AI-generated output for public verification.
Our operating principles
Control, transparency and proportional claims.
Trust is not a badge. It comes from the way capabilities are designed, documented and represented to the people using them.
Keep humans in control
Scripts, voices, assets, dynamic records, mix settings and delivery selection remain configurable by the user.
State the limitation
Experimental capabilities and inconclusive detector results are described as such instead of being presented as certainty.
How AI is used
AI assists several stages. It does not hide the workflow.
Different features use AI for different purposes. The production workspace remains the place where the resulting script, assets and sound are reviewed together.
Generate a starting direction, not an irreversible answer.
Voxify can analyze a supplied URL or product brief, propose scripts, rephrase an existing script, add performance cues and recommend matching voice and music options.
Voice and sound assets stay identifiable in the project.
Generated voice-over and sound-effect assets are placed in the same timeline as licensed library assets and user uploads.
Mastering can be previewed A/B.
AI Mastering is an optional finishing action. Users can compare the processed and unprocessed result before export.
Issue spotting is experimental, not legal approval.
The Compliance Check can flag possible disclosure, claim or brand-safety concerns. It is explicitly presented as a preview and does not replace qualified legal or regulatory review.
Voice policy
Voxify does not currently offer voice cloning.
The voices available in the product are licensed catalog voices or voices generated from a written description. They are not created from a recording of an identifiable person supplied for cloning.
- Uploaded recordings can be imported, transcribed, separated and mixed.
- Uploaded voice recordings are not used by Voxify to build a voice model or voiceprint.
- The person uploading audio must have the rights and consents needed to use it.
- If voice cloning is introduced in the future, the policy will be updated before launch.
Article 50(2)-aligned marking
Every final AI-generated audio variant carries two marking layers.
Voxify embeds an imperceptible watermark in the final audio signal and writes machine-readable origin metadata on the deliverable. A public detector reports the two layers independently.
- The watermark is embedded in the final mix before encoding.
- Metadata provides generator, content-type and digital-source descriptors.
- Recognized provenance may list voice and music assets.
- A missing mark or negative result is not proof of human origin.
The two-layer design supports the machine-readable marking and detectability requirements in Article 50(2) of the EU AI Act. It does not replace the context-specific obligations of publishers or deployers.
Verification layersReported separately
Audio watermarkImperceptible marking checked in the audio signal
PrimaryMetadata markingGenerator and content descriptors where preserved
DetailProvenance lookupVoice and track data for a recognized rendering
AssetsExplicit caveatNegative results remain inconclusive
ContextData & infrastructure
The practical information procurement teams need.
The complete legal terms remain the authoritative source. This summary links directly to the current policies instead of replacing them.
Hosting location
Application hosting, database and object storage currently operate on AWS in US East (N. Virginia). International transfer safeguards are described in the Privacy Policy.
Review transfers →Security measures
Published measures include encryption in transit, hashed passwords, access controls and session management. No online service can promise perfect security.
Review security →Service providers
The current sub-processor list covers AI and speech providers, infrastructure, payments, analytics, CDN services and licensed music sources.
Review sub-processors →Processor relationships
For API and Embedded customers, Voxify can process end-user data on the customer's behalf under the published Data Processing Addendum.
Read the DPA →Known boundaries
What the platform does not claim.
No universal AI detectorThe public detector verifies Voxify marking only; it cannot classify all audio from every provider.
No automatic legal approvalCompliance Check is an experimental issue-spotting aid and not legal or regulatory advice.
No current voice cloningVoxify does not currently create a synthetic copy from a real person's supplied recording.
Need a direct answer for your review?
Ask about data handling, sub-processors, integration architecture or the capabilities available today.