Inaccurate results from 0.9.3 model

I have recording in mp3 format which i converted into wav using ffmpeg and it sounds absolutely clear.

!ffmpeg -i result1.wav -ar 16000 output1.wav

On using the deepspeech model- i am getting absurd response. following is my code-

from deepspeech import Model
import wave
import numpy as np
model = Model(’./deepspeech-0.9.3-models.pbmm’)
fin =‘result1.wav’, ‘rb’)
audio = np.frombuffer(fin.readframes(fin.getnframes()), np.int16)

Perform inference

infered_text = model.stt(audio)

Please suggest if i have missed anything or let me know for further information.

Hello! Thanks so much for getting in touch and I’m so sorry that I’m only going to be able to offer very limited support. The Common Voice is a Mozilla Foundation project that is an opensourced, multilingual speech corpus and the platform and tools dedicated to supporting and collecting this data. I’m so sorry that Deepspeech is a separate project held by the Mozilla Corporation and I’m not clear on how actively it is being maintained and supported.

My expertise and the scope of this forum is limited to the Common Voice datasets, perhaps the Deepspeech Github might be a better source of information?