Legal / Attribution

Third-party notices

Srutam's on-device speech recognition is built from the following components.

Speech recognition model: NVIDIA Parakeet-TDT-CTC-110M

Creator: NVIDIA. Source: NVIDIA Parakeet-TDT-CTC-110M on Hugging Face.
License: Creative Commons Attribution 4.0 International (CC BY 4.0).
Changes: the model was converted to ONNX and quantized to int8 by the sherpa-onnx project (ASR model releases, file sherpa-onnx-nemo-parakeet_tdt_ctc_110m-en-36000-int8). Srutam uses it unmodified from that export.
NVIDIA does not endorse Srutam.

Voice activity detection model: Silero VAD

Source: Silero VAD on GitHub.
License: MIT

Inference runtime: sherpa-onnx and ONNX Runtime

sherpa-onnx: Apache License 2.0. Prebuilt libsherpa-onnx-jni.so and the Kotlin wrappers under app/src/main/java/com/k2fsa/sherpa/onnx/ come from release v1.12.39.
ONNX Runtime: MIT License