VoxRT ASR streaming model weights
Pre-compiled, encrypted streaming speech-recognition model weights in VoxRT's .vxrt format, for consumption by the VoxRT on-device runtime on Android and iOS.
This repository distributes pre-compiled streaming-medium-pc automatic-speech-recognition model weights in VoxRT's own .vxrt format, packaged for the VoxRT on-device inference runtime and paired with separate Android and iOS consumer libraries. A .vxrt file is a self-contained, binary, framework-agnostic model file with topology and weights in one compact format, encrypted at rest with AES-256-GCM and decrypted on load by the runtime, with a version compatibility table since newer files may use features an older runtime cannot load.
The repository itself is a thin index; the actual weight files are attached to GitHub releases rather than committed to the git tree, and are downloaded and checksum-verified separately. The encoded weights are derived from NVIDIA's stt_en_fastconformer_hybrid_medium_streaming_80ms_pc model, released under CC-BY-4.0, re-packaged into the .vxrt format with fp16 weights and no fine-tuning or architectural changes; attribution to NVIDIA is required on redistribution. The VoxRT runtime that consumes the file is separately licensed and proprietary.