Testing Mozilla DeepSpeech

These are the steps to test deepspeech 0.4.1 on Linux Ubuntu 18.04:

Install libsox and fake it as libsox2:

$ sudo apt-get install libsox-dev

$ ln -s libsox.so.3.0.0 libsox.so.2

Record your voice:

$ rec -c 1 -r 16000 -b 16 recording.wav

Alternatively download the audio from google translator and convert to wav 16bit 16khz mono:

$ ffmpeg -i the_quick_fox_jump_over_the_lazy_dogs.mp3 -acodec pcm_s16le -ac 1 -ar 16000 recording.wav

Test it:

$ ./deepspeech --model ../deepspeech-0.4.1-models/models/output_graph.pbmm --alphabet ../deepspeech-0.4.1-models/models/alphabet.txt --lm ../deepspeech-0.4.1-models/models/lm.binary --trie ../deepspeech-0.4.1-models/models/trie --audio recording.wav
TensorFlow: v1.12.0-10-ge232881
DeepSpeech: v0.4.1-0-g0e40db6
the quick brown fox jump over the lazy dog

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s