Basis AI released Basis Global Emotions on October 12, 2026, saying it provides 525.17 hours of custom-recorded speech from 10,420 speakers across 108 locales. It is the company's first major release in the speech dataset category since its previous work on Basis Conversations 1500.
Basis AI reported 525.17 hours of audio recordings, measured across 137,696 matched series and 54,575 unique speakers. That compares with its earlier release, Basis Conversations 1500, which contained 1,500 hours of multilingual, multi-party conversation.
Basis Global Emotions is built on native speaker recordings and targets expressive speech analysis across languages and regions. Availability begins on Hugging Face, initially for researchers and developers.
"From a neutral reading to full frustration, or from ordinary speech to a whisper, Basis Global Emotions covers a wide range of emotional expressions," said Alexis Sursock, community article author. The dataset supports studying changes in rhythm, pitch, emphasis, and delivery.
The announcement follows the release of Basis Conversations 1500. Basis AI said the two datasets offer different ways to study the range of human speech, with one focusing on spontaneous interaction and the other on expressive speech.
Basis AI did not say whether the dataset includes human-validated transcriptions, and noted that transcripts are not included in this release. The company said researchers can use the dataset to train models with more control over delivery.
Source: huggingface
