Nickolay V. Shmyrev

Update to tensorflow 1.0.0

1

Fixes issue #40, issue #41, issue #36

CUDA online decoder requires exactly frames_per_chunk samples

5

It is not documented or enforced in the API but actually cuda online pipeline requires audio chunk to have exactly GetNSampsPerChunk of audio data. There is no buffering in cuda...

Very slow xvector computation with all time spend on compilation

11

While running Voxceleb with different architectures I noticed that xvector extraction is very slow: ``` nnet3-xvector-compute --verbose=0 --use-gpu=no --min-chunk-size=25 --chunk-size=10000 \ --cache-capacity=64 "nnet3-copy \ --nnet-config=exp/xvector_nnet_1a/extract.config \ exp/xvector_nnet_1a/final.raw - |" "ark:apply-cmvn-sliding...

Support for Vosk speech recognition server over websocket protocol

9

For more details on Vosk see https://github.com/alphacep/vosk-server https://github.com/alphacep/vosk-api

Training and decoding speed

1

Hi Thank you for the nice software. Could you please share the information: 1) How long the training on librispeech takes on how many GPUs 2) How fast is the...

Fix python method naming

1

Should be lowercase actually

0.3.42 works, 0.3.43 doesn't 

Unity Support

24

There are projects like https://github.com/Babilinski/vosk-stt-unity and https://github.com/Yeti47/Vosk4Unity Eventually we need to have our own project for that.

Integrate Rust bindings

6

https://github.com/wzhd/vosk-rs and https://github.com/shakacode/vosk-rs

Setup fine-tuning script for the models

28

As in https://github.com/daanzu/kaldi-active-grammar/issues/33 https://github.com/gooofy/zamia-speech/issues/106