Humaneness Voice 1.1 Alpha

LAION’s Humaneness Voice 1.1 Alpha repositories provide experimental large and small speech models while training remains in progress. The English cards describe synthesis from literal text with optional performance directions and a short reference recording or trained voice alias. Both repositories contain interim model weights, inference code and configuration, and link to a public listening comparison. LAION says quality evaluation is incomplete and recommends consent and disclosure when generating speech; the published checkpoint and demo do not establish a quality improvement or deployment readiness.