Open In App

Python | Speech recognition on large audio files

Last Updated : 23 Jul, 2019
Like Article

Speech recognition
is the process of converting audio into text. This is commonly used in voice assistants like Alexa, Siri, etc. Python provides an API called SpeechRecognition to allow us to convert audio into text for further processing. In this article, we will look at converting large or long audio files into text using the SpeechRecognition API in python.

Processing Large audio files

When the input is a long audio file, the accuracy of speech recognition decreases. Moreover, Google speech recognition API cannot recognize long audio files with good accuracy. Therefore, we need to process the audio file into smaller chunks and then feed these chunks to the API. Doing this improves accuracy and allows us to recognize large audio files.

Splitting the audio based on silence

One way to process the audio file is to split it into chunks of constant size. For example, we can take an audio file which is 10 minutes long and split it into 60 chunks each of length 10 seconds. We can then feed these chunks to the API and convert speech to text by concatenating the results of all these chunks. This method is inaccurate. Splitting the audio file into chunks of constant size might interrupt sentences in between and we might lose some important words in the process. This is because the audio file might end before a word is completely spoken and google will not be able to recognize incomplete words.

The other way is to split the audio file based on silence. Humans pause for a short amount of time between sentences. If we can split the audio file into chunks based on these silences, then we can process the file sentence by sentence and concatenate them to get the result. This approach is more accurate than the previous one because we do not cut sentences in between and the audio chunk will contain the entire sentence without any interruptions. This way, we don’t need to split it into chunks of constant length.

The disadvantage of this method is that it is difficult to determine the length of silence to split because different users speak differently and some users might pause for 1 second in between sentences whereas some may pause for just 0.5 seconds.

Libraries required

Pydub: sudo pip3 install pydub
Speech recognition: sudo pip3 install SpeechRecognition


Input:  peacock.wav 


exporting chunk0.wav
Processing chunk 0
exporting chunk1.wav
Processing chunk 1
exporting chunk2.wav
Processing chunk 2
exporting chunk3.wav
Processing chunk 3
exporting chunk4.wav
Processing chunk 4
exporting chunk5.wav
Processing chunk 5
exporting chunk6.wav
Processing chunk 6


# importing libraries
import speech_recognition as sr
import os
from pydub import AudioSegment
from pydub.silence import split_on_silence
# a function that splits the audio file into chunks
# and applies speech recognition
def silence_based_conversion(path = "alice-medium.wav"):
    # open the audio file stored in
    # the local system as a wav file.
    song = AudioSegment.from_wav(path)
    # open a file where we will concatenate  
    # and store the recognized text
    fh = open("recognized.txt", "w+")
    # split track where silence is 0.5 seconds 
    # or more and get chunks
    chunks = split_on_silence(song,
        # must be silent for at least 0.5 seconds
        # or 500 ms. adjust this value based on user
        # requirement. if the speaker stays silent for 
        # longer, increase this value. else, decrease it.
        min_silence_len = 500,
        # consider it silent if quieter than -16 dBFS
        # adjust this per requirement
        silence_thresh = -16
    # create a directory to store the audio chunks.
    # move into the directory to
    # store the audio files.
    i = 0
    # process each chunk
    for chunk in chunks:
        # Create 0.5 seconds silence chunk
        chunk_silent = AudioSegment.silent(duration = 10)
        # add 0.5 sec silence to beginning and 
        # end of audio chunk. This is done so that
        # it doesn't seem abruptly sliced.
        audio_chunk = chunk_silent + chunk + chunk_silent
        # export audio chunk and save it in 
        # the current directory.
        print("saving chunk{0}.wav".format(i))
        # specify the bitrate to be 192 k
        audio_chunk.export("./chunk{0}.wav".format(i), bitrate ='192k', format ="wav")
        # the name of the newly created chunk
        filename = 'chunk'+str(i)+'.wav'
        print("Processing chunk "+str(i))
        # get the name of the newly created chunk
        # in the AUDIO_FILE variable for later use.
        file = filename
        # create a speech recognition object
        r = sr.Recognizer()
        # recognize the chunk
        with sr.AudioFile(file) as source:
            # remove this if it is not working
            # correctly.
            audio_listened = r.listen(source)
            # try converting it to text
            rec = r.recognize_google(audio_listened)
            # write the output to the file.
            fh.write(rec+". ")
        # catch any errors.
        except sr.UnknownValueError:
            print("Could not understand audio")
        except sr.RequestError as e:
            print("Could not request results. check your internet connection")
        i += 1
if __name__ == '__main__':
    print('Enter the audio file path')
    path = input()

Output :


The peacock is the national bird of India. They have colourful feathers, two legs and 
a small beak. They are famous for their dance. When a peacock dances it spreads its 
feathers like a fan. It has a long shiny dark blue neck. Peacocks are mostly found in 
the fields they are very beautiful birds. The females are known as 'Peahen1. Their 
feathers are used for making jackets, purses etc. We can see them in a zoo. 

Like Article
Suggest improvement
Share your thoughts in the comments

Similar Reads