Tenstorrent and Smallest.ai offer real-time text-to-speech on on-premises hardware

Tenstorrent

Tenstorrent and Smallest.ai partnered to run Smallest.ai's Lightning V2 real-time text-to-speech model natively on Tenstorrent hardware, for deployments on a customer's own servers. The stack runs on Tenstorrent Galaxy Blackhole servers and is available now on usage-based pricing with no upfront commitment. The companies target financial services, telecom, healthcare and enterprise customers that need their data kept under their control.

• Smallest.ai's engineer claims 4x gains at a fraction of the cost of comparable GPU infrastructure; the release gives no cost, latency or concurrency figures.
• The release covers text-to-speech only; speech recognition, speech-to-speech and a full voice agent stack are not part of it.
• The release emphasises full data sovereignty but names no compliance certifications.
• Lightning V3 is not included in this release; the companies say work to bring it to Tenstorrent is underway.