I am doing the Japanese Language proficiency test and it has a listening component. Some listening exercises are long ones and when practicing I don't understand the parts when needed. I'd like to automatically split practice recordings into phrases or sentences to tracks. I want to avoid listening (or reading its transcript) to the recording (number of times) as much as possible so standard software that can slow it down for splitting is not ideal or autotranscription and then autoreproduction (which probably gives somewhat unnatural speech anyway) is not great Is there an automatic way to do this?


Unfortunately there isn't a way to do this automatically.


Assuming there are moments of silence between sentences or exercises, if you define the dB of silence a program like audacity can split the audio file into segments whenever the sound drops below the specified dB level for a duration of time that you specify (.5 seconds, 2 seconds etc.). You'll want to know that the dB for silence and voice are consistent across the audio.

The question is answered in stack overflow here.

  • This does not answer the OP's question. – Mark Apr 28 '20 at 4:58
  • OP asks for a way to split audio tracks by phrase or sentence. Sentences (and long phrases) often end with an audible pause/silence. Splitting at those points means splitting audio tracks by sentences. Am I misunderstanding a key detail in the question? – Mark K Apr 28 '20 at 10:58
  • The OP wants to split by sentence or phrase. Using the assumption that there are moments of silence between sentences or exercises doesn't assist as it is not going to achieve the OP's end goal. Unless a very complex AI is used to determine phrasing or sentence structure, simply using strip silence won't achieve the OP's requirement. – Mark Apr 28 '20 at 14:42
  • @Mark I defer to your experience, as I have not done this myself. The OP asking for how to approach an answer. I'm not guaranteeing success but giving a possible solution. However, this site seems to agree with me that you can use strip silence to split spoken audio by sentences. (See second section) logicpro.skydocu.com/en/create-a-song-arrangement/… – Mark K Apr 28 '20 at 16:09
  • It's wrong. the problem is that you cannot assume anything about the reason why someone might pause during a sentence, or a phrase, so simply assuming that a 'gap' has some sort of equivalence to a 'phrase' or a 'sentence' is a fallacy - an incorrect assumption. IF you were to go down this path, you would have to clearly define sentences or phrases in terms of measurable gaps - and that is not possible to do without a good deal of work, which the OP is attempting to avoid having to do. The other issue with this question is that it relates more to consumer use of audio and is therefore O/T. – Mark Apr 29 '20 at 1:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.