P

Parakeet Tdt Ctc 0.6b Ja

Developed by nvidia
Parakeet TDT-CTC 0.6B is an automatic speech recognition (ASR) model capable of transcribing Japanese speech with punctuation, developed by the NVIDIA NeMo team.
Downloads 4,184
Release Time : 5/13/2024

Model Overview

This model is an XL version of the hybrid FastConformer TDT-CTC architecture, specifically designed for Japanese speech recognition tasks, capable of handling speech transcription with punctuation.

Model Features

Hybrid architecture
Combines FastConformer and TDT-CTC architectures to optimize speech recognition performance
Efficient inference
The TDT architecture significantly improves inference speed by decoupling token and duration prediction
Japanese language support
Specifically optimized for Japanese speech recognition, supporting transcription with punctuation
Large-scale training
Trained on over 35k hours of Japanese speech data

Model Capabilities

Japanese speech recognition
Punctuation transcription
16kHz mono audio processing

Use Cases

Speech transcription
Japanese speech to text
Convert Japanese speech content into text with punctuation
Achieves a CER of 6.4% on the JSUT basic5000 test set
Speech content analysis
Analyze and process Japanese speech content
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
Š 2025AIbase