Bad transcription from microphone

Archives DeepSpeech
#1

hello i’am using mic_vad_streaming.py with french model. I 'am using this command

python mic_vad_streaming.py -m fr.pbmm -s fr.scorer -v 0 -d 1 -r 16000 -w /audio

i have recorded “un deux trois quatre cinq six sept huit neuf dix” the results is “un de trois quatre cinq six c’est une dix”

As i have saved the audio. I 'have tested the offline version that has given also “un deux trois quatre cinq six sept huit neuf dix”

Any help please ?

#2

Can you give some more information about your setup and share the audio file and where you got the models from? Also you can join us for realtime support on Mozilla’s Matrix.