Speech: what I’ve logged
Speech comes up in 2 posts of my AI log, from October 2026 to October 2026, and each item below links to the primary source. The sources come from NVIDIA and Block, filed under frontier, open weights and science.
Items that mention Speech
- NVIDIA 28 Oct 2026
NVIDIA Nemotron 3.5 Transcribe and Granary v2
Open-weight promptable speech model covering transcription, speaker labels, translation and audio Q&A in 40 languages, plus a 463,062 hour multilingual dataset release.
- Block 23 Oct 2026
Dessa: a conversation model trained only on speech
A full-duplex speech-to-speech model trained from scratch on human conversation with no text, said to run faster than real time and locally on an M4 Max MacBook Pro or better.