audio.cpp Adds Four New ASR Models

Acceptable-Cycle4645 · reddit · 2026-07-09

audio.cpp has added 4 ASR models: Nemotron 3.5 ASR, Higgs Audio STT, VibeVoice ASR, and Hviske ASR, along with initial streaming support. The author notes these implementations can be 1.07x to 2.41x faster than their Python counterparts.

Original post →

More from Infra

Infra channel →