The service
Audio Lab is provided free of charge, as-is, with no account required. We don't guarantee that every tool will be available, error-free, or fit for a particular purpose at all times, and we don't offer a warranty on the results a tool produces. Processing runs in your own browser, so performance and the exact results you get can vary with your device, browser, and available memory; we don't promise identical output across every device or browser version. You use Audio Lab at your own discretion, and for anything you depend on for a real deadline or a paid deliverable, it's worth double-checking the output before relying on it.
Your content
Audio you open, record, or generate using Audio Lab stays yours. We don't claim any ownership over it, and since it's never uploaded to us, we don't have a copy of it to claim ownership over in the first place.
You're responsible for having the right to process whatever you feed into a tool. That includes copyrighted music you don't have the rights to distribute or remix, and it includes other people's voices: cloning, tuning, or otherwise manipulating a voice that isn't yours requires that person's consent. Using Audio Lab to process content you don't have the rights to is your responsibility, not ours.
Voice cloning and generated voices
Voice cloning and other voice-generation tools are meant for uses you have a legitimate right to: your own voice, a voice you have explicit permission to use, or clearly fictional or consented material. Using these tools to impersonate a real person without their consent, to commit fraud, to harass someone, or to create deceptive deepfakes intended to mislead people about who is actually speaking, is not an acceptable use of Audio Lab, regardless of how the output is eventually used.
Open-model licenses
The AI tools on Audio Lab run on open-weight models, each under its own open-source license. Those licenses govern the models themselves and, depending on the license, may have implications for how you use their output — it's worth checking the specific license if you plan to use AI-generated audio commercially.
| AI tool(s) | Model | License |
|---|---|---|
| Stem splitter, AI-mode vocal remover | Demucs v4 | MIT |
| Speech to text | Whisper | MIT |
| Text to speech | Kokoro-82M | Apache-2.0 |
| Voice cloning | Chatterbox | MIT |
Acceptable use
Beyond the voice-specific rules above, please don't use Audio Lab to:
- Infringe someone else's copyright or other intellectual property rights.
- Impersonate a real person, organization, or brand in a way meant to deceive.
- Harass, defame, or threaten another person.
- Generate or distribute deliberately misleading audio, including deceptive deepfakes.
- Attempt to disrupt, overload, or reverse engineer the service in bad faith.
Limitation of liability
Audio Lab is offered as a free tool, and to the extent the law allows, we're not liable for losses or damages arising from your use of it — including lost audio, a missed deadline, or a result that didn't match what you expected. This extends to output from the AI models the tools rely on: transcripts, cloned voices, and separated stems can all contain errors, and we're not liable for decisions made based on inaccurate AI output. If you're using Audio Lab for something high-stakes, keep your own backups and verify your own results before relying on them.
Changes to these terms
We may update these terms from time to time, and we'll update the "Last updated" date at the top of this page when we do. Continuing to use Audio Lab after a change means you accept the updated terms.
Contact
Questions about these terms can go to the contact address published on this page.