---
title: "Ethical AI voice and music tools: consent, cloning and artist rights | Ethical Stack"
url: https://ethicalstack.ai/guides/ai-voice-consent/
content_signal: search=yes, ai-input=yes, ai-train=no
---

# Whose voice is your AI using?

A convincing voice tells you little about the permission behind it. Check the original training material, the voice or music you add, and what happens to recordings afterward.

## Permission for training and cloning

Look for the source of the model’s speech recordings and, for music, both recordings and compositions. A voice-cloning consent check covers the added voice; it does not clear every source used to train the base model. Check whether permission covers the intended use and whether the performer is paid.

Our [training-data rubric](https://ethicalstack.ai/methodology/#training-rights) credits documented permission and creator controls. [Fairly Trained’s requirements](https://www.fairlytrained.org/certifications) cover training sources, inherited models and synthetic data; any certification needs to match the specific model or service you are considering.

## Keeping control of a voice

Check how a service verifies the speaker, who can use a shared voice, and whether access can be revoked. Deleting an uploaded recording, removing a voice from a library and undoing earlier model training are different actions. Where a provider does not explain what deletion covers, the profile records that gap.

## Recordings, transcripts and calls

Check participant consent, recording defaults, who can read transcripts, retention and reuse for training. Business APIs and consumer apps can have different terms. Running speech recognition locally can avoid a cloud upload, but connected apps and backups still matter.

For a voice agent, also check whether callers know they are speaking to AI, how to reach a person, and what actions the agent can take.

## Compare tools for your task

[Generate or dub a voice](https://ethicalstack.ai/compare/voice-generation/) · [Transcribe audio](https://ethicalstack.ai/compare/transcription/) · [Make music and sound](https://ethicalstack.ai/compare/music-generation/) · [Build a voice agent](https://ethicalstack.ai/compare/voice-agents/)

Profiles separate product controls, model training evidence and the hosting route. Start with the task, then open the findings that matter to you.