rnnt

PyTorch implementation of "Transformer Transducer: A Streamable Speech Recognition Model with Transformer Encoders and RNN-T Loss" (ICASSP 2020)

A curated list of awesome papers on contextualizing E2E ASR outputs

An efficient implementation of RNN-T Prefix Beam Search in C++/CUDA.

An implementation of RNN-Transducer loss in TF-2.0.

FunASR实时语音识别版，识别麦克风和电脑内播放的声音，电脑语音打字软件

I'm building an end-to-end Vietnamese Speech Recognition System. I'll deploy it into production with the help of Flask, Uwsgi, Nginx, and AWS ...

Conformer RNN-Transducer

NVIDIA Nemotron 3.5 ASR (multilingual streaming, 600M) exported to ONNX: cache-aware encoder/decoder/joiner + a numpy+onnxruntime streaming engine. WER 0.0137 vs original, 3.8x realtime on CPU.

Pure PyTorch implementation of the loss described in "Online Segment to Segment Neural Transduction" https://arxiv.org/abs/1609.08194

🔊 Enhance speech recognition with GLM-ASR-Nano-2512, a high-performance model excelling in dialect support and low-volume audio accuracy.

Local-first Vietnamese/English streaming speech-to-text PWA with private voice personalization

Deep learning-based subtitle generation model that processes audio datasets to generate accurate text transcriptions. Includes audio feature extraction, encoder-decoder architecture, training pipelines, and evaluation metrics for subtitle alignment.

NVIDIA Conformer-CTC and Conformer-Transducer (RNN-T) running natively on Apple Silicon via MLX. Loads NeMo checkpoints directly.

🚀 Create and manage SPL tokens on the Solana blockchain with ease, using our Next.js-based launchpad for streamlined token and liquidity management.