Kled just entered voice.

Users can record and upload real-world speech, capturing tone, accent, emotion, and context in a way models can actually learn from.

We’ll initially support multilingual uploads across: Malay (Bahasa Malaysia), Indonesian (Bahasa Indonesia), and Filipino/Tagalog.

Voice is one of the most under-collected, high-value data types in AI today. Accents, tone, emotion, environment. This is the layer models still struggle with.

All voice data will be packaged and sold to leading voice foundation model teams, both directly and through our distribution partners.

We’ve designed 50 starter tasks, rolling out in our V3 update, so users can start contributing immediately.