Block’s Dessa: The Speech AI That Never Touched Text

Block has introduced Dessa, a full-duplex speech model trained solely on human audio with no text data and no pretrained weights, according to byteiota. The model is positioned for developers building voice applications.

Block's Dessa is a full-duplex speech model trained solely on human audio — no text, no pretrained weights. Here's what it means for developers building voice applications. The post Block’s Dessa: The Speech AI That Never Touched Text appeared first on byteiota .