Specifications
Size
~350 MB (INT8 ONNX)
Architecture
Conformer
Latency
Medium
Language
English, Spanish, German, French + translation
Developer / Creator
NVIDIA (NeMo team), Sherpa ONNX community
Download Source
Verified Repository Source
Hugging Face Hub / Sherpa ONNX model catalog
Open Model Repository (k2-fsa/sherpa-onnx)Model Overview
NVIDIA's Canary is an advanced multi-lingual speech-to-text and translation model. It supports English, Spanish, German, and French speech recognition, and can transcribe and translate between these languages on-device. It runs locally in tapWhisper using Sherpa ONNX with high efficiency.
Available Model Variants
| Model Name | File Size | RAM Usage | Format/Quant | Languages | Description |
|---|---|---|---|---|---|
| NVIDIA Canary | 350 MB | 650 MB | INT8 (ONNX) | EN, ES, DE, FR | NVIDIA Canary 180M Flash. Supports on-device ASR and speech translation. |