W4D/parakeet-tdt-0.6b-v3-onnx

parakeet-tdt-0.6b-v3 ONNX

ONNX export of nvidia/parakeet-tdt-0.6b-v3 for use with onnx-asr.

Multilingual speech-to-text model supporting 25 European languages with automatic language detection.

Usage

import onnx_asr

model = onnx_asr.load_model("nemo-conformer-tdt", "path/to/this/repo")
print(model.recognize("audio.wav"))

Files

File	Size	Description
encoder-model.onnx	~2.4 GB	FastConformer encoder
decoder_joint-model.onnx	~69 MB	TDT decoder + joint network
vocab.txt	~100 KB	Tokenizer vocabulary (8193 tokens)
config.json	~124 B	Model configuration

Requirements

pip install onnx-asr[cpu]

Supported Languages

Bulgarian, Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, German, Greek, Hungarian, Italian, Latvian, Lithuanian, Maltese, Polish, Portuguese, Romanian, Russian, Slovak, Slovenian, Spanish, Swedish, Ukrainian.