Tether has released version 0.14.0 of its QVAC SDK, updating its local AI development stack with a focus on faster on-device inference for mobile, expanded multilingual speech, and cross-platform compute adaptation.
According to Odaily, the update adds an OpenClaw-compatible path intended to help developers deploy local coding agents at lower cost.
The Supertonic3 speech component expanded its text-to-speech language support to 31 languages from five. Chatterbox and Supertonic now support Android GPU operation and include memory usage optimizations.
At the system level, Linux adds a dynamic compute backend that can automatically match the optimal execution environment while a program is running. The architecture is described as technical groundwork for future compatibility with CUDA and ROCm GPU computing, aimed at lowering the deployment threshold for local multimodal AI projects.