Introducing **Eleven Eleven**, a family of datasets constructed using ElevenLabs models. We provide extensive collections of voices and sound effects, including **550K sound effects** and **16K unique voices across 90+ languages**.

## Reasoning Behind

Sleeping AI is releasing these datasets as high-quality training materials for training state-of-the-art models and for sharing with trusted partners.

Trusted partners are organizations and individuals we collaborate with and trust to handle the database responsibly, without surfacing or redistributing it online without the explicit knowledge of Sleeping AI.

If we believe Sleeping AI's terms and conditions have been breached, we reserve the right to publicly identify the organizations we have shared the data with. We will not take responsibility for violations committed by third parties.

## Are We Not Doing Open Source?

Maybe.

We still release early versions to trusted partners and share data openly whenever we believe doing so will not violate applicable terms and conditions. There are limits to what can reasonably be pursued or enforced.

We believe it is better for teams to have access to useful research materials than for nobody to have access to them.

## Data Governance

We have **zero tolerance for breaches of our terms and conditions**. Organizations or individuals found to have violated them may be publicly named, as Sleeping AI will not take the fall for someone else's actions.

We also ensure that all data shared under **Project Imagine** is created using commercially available state-of-the-art models.

## Funding

We thank all of our funders and supporters who help Sleeping AI pursue and contribute to open research and open-source initiatives.