What's a good, ideally free workflow to record bunch of audio files?

April 24, 2019 11:38 AM

Let's say I have 500 words in a spreadsheet. I want an audio file per word, numbered, in order. What's the best way to do that, resilient to errors?

For smaller numbers of words I've simply left a 3 second space between each entry, then used audacity to detect silence and split. The annoyance is errors in recording, which get quite likely as the number of entries go up. So I'm wondering if there's a better workflow to go about this. Ideally sit down, record them in order, then painlessly get numbered audio files...and resilient to the occasional error.

posted by wooh to Computers & Internet (7 answers total)

Question: Does the spoken word need to be performed by a human? There are plenty of ways to use text-to-speech to read a word into a file...
posted by Wild_Eep at 11:40 AM on April 24, 2019

Simple refinement: Clap after you make an error. On the waveform, it'll be obvious which ones to discard.

In the past I've done this with a custom Python script using pyaudio. If you have the baseline skills, this will get you most of the way there.
posted by supercres at 11:51 AM on April 24, 2019 [2 favorites]

Of course, you don't really need to detect, which is the fiddliest bit. You could just print the word, start recording, then trim [if you want] and save, maybe with playback and an input for whether to continue or rerecord.
posted by supercres at 11:55 AM on April 24, 2019

Interesting!

Here’s another method of doing this with python. You could also combine this with a list containing each word in order to assign these files the word as a file name.

(You may want to have a method of dealing with mistakes, such as doing 100 words at a time instead of 500, to save you possible headaches..)
posted by elephantsvanish at 11:55 AM on April 24, 2019

Response by poster: Relevant details: I am a programmer, just not super familiar with audio stuff. So python related solutions are welcome. I am sort of hitting myself for not thinking of that...it could prompt them for each word or something...

Human speech is necessary (can't use text to speech).
posted by wooh at 12:15 PM on April 24, 2019

Yep, I'd do something like this pseudocode:

for word in words:
    print word
    start recording
    wait N seconds
    stop recording
    play back recording
    prompt for input
    if input = 'q':
        rerecord
    elif input = 'p':
        save file
        [optionally: trim based on threshold]
        go to next word


Probably a "while" in there for each word in the list, like "done = False; while not done:"

Almost exactly what I did before if memory serves.
posted by supercres at 3:12 PM on April 24, 2019

Oh, check for presence of that word's file first. If they go through them all and want to rerecord a specific word, just delete that file.
posted by supercres at 3:14 PM on April 24, 2019