Voice is the most personal asset you have.

1 · Who can be cloned

• Yourself, for any purpose.
• People on your team who have signed a written voice-license agreement with you.
• Talent you have hired with explicit, written voice-use rights.
• Voices already in the public domain (e.g. LibriVox recordings, government-released audio).

We will not help you clone:

• Celebrities, public figures, or public-domain personalities without their explicit written permission.
• Friends, family members, or coworkers who haven't signed off in writing.
• Voices scraped from YouTube, TikTok, podcasts, or interviews without the speaker's permission.
• Anyone living or deceased for the purpose of misrepresenting them, defrauding their family, or impersonating them in any communication.

2 · How training data is handled

Bloom Voice runs on your hardware. There is no upload step, no Bloom server in the loop, and no telemetry channel for audio.

• Training audio stays on the machine you installed Bloom Voice on. It is never copied to Bloom infrastructure.
• Generated audio stays on your machine. We do not maintain a copy.
• Voice profiles are local files. Move them, back them up, or delete them — same as any other file in your project.
• The hosted demo on bloom-voice.pages.dev is the one exception: demo audio you submit is destroyed within 24 hours and is never used as training data for any model.

3 · Commercial use

Once you have purchased the Bloom Voice code drop, the code is yours to ship inside any commercial product, with two caveats:

• The licensing of any output audio is your responsibility. If you’re cloning a hired voice actor, make sure your contract with them covers commercial use of synthetic audio.
• You may not relicense, repackage, or resell the Bloom Voice source code itself as a competing voice-cloning product or service.
• Internal tools, client products, customer-facing apps, and audiobooks built on top of Bloom Voice are all explicitly fine.

4 · Required consent flow

Bloom Voice ships with a built-in consent prompt that fires the first time a new voice profile is trained. The prompt asks you to confirm one of:

• This is my own voice.
• I have written permission from the speaker, attached or referenced.
• This recording is in the public domain.

The prompt is not removable in the standard distribution. If you need a different flow for enterprise compliance, contact us directly.

5 · Reporting violations

If you believe a Bloom Voice user has cloned a voice without consent, email hello@thaliabloom.com with as much context as you can share. We can’t revoke already-distributed code, but we can refuse future sales, disconnect support, and document the case publicly when appropriate.

6 · What we will never do

• Add a phone-home telemetry channel to the code drop.
• Add a license-check that calls our servers.
• Add an automatic update path that lets us change behavior on your machine.
• Sell your audio, your voice profiles, or your generated output to anyone.
• Train our own model on customer audio.