Locating chapter breaks in a recording

OS: Windows 7 Pro SP1, 64-bit
Audacity: 3.0.3, 64-bit

Project: I’m digitizing Dad’s audio books from cassettes to MP3 files of 70-75 minute ‘parts’. 70-75 minutes are manageable on his iPod and they can also be burned to CDs.

Problem: I need some way to detect chapter breaks in Audacity’s audio recording. The chapter breaks are basically a few seconds of leading silence, chapter announcement, a few seconds of trailing silence. An example is shown below.

ChapterSound.jpg
I’ve played with Analyze > Label Sounds… and can’t figure out how to isolate the chapter breaks. I think I need a Label Sound specified something like this: Leading silence = 2.5 seconds, sound length = 0.5 seconds, trailing silence = 1.0 seconds. Note: I think v3.0.2 had an Analyze > Sound Finder… tool, but it doesn’t seem to be in v3.0.3.

Anybody know how I can get Audacity to do this work for me? It would be ideal if Audacity located ONLY the chapter breaks (e.g. 4-5 chapter breaks in a 75 minute recording).

Thanks,
Rose

How many seconds’?

chapter announcement,

How long is that announcement?

a few seconds of trailing silence.

How many seconds’?

Thank you for the fast reply!

Hard questions to give a firm answer because every audiobook is different. Even within the same audio book the chapter breaks vary a little.
In this example:

Leading silence = 1.30 seconds.
Announcement = 0.35 seconds. This announcement is “two”, but I’ve also found announcement of “Chapter two”.
Trailing silence = 2.30 seconds.

Hope this helps.
Thanks,
RoseM

Use the “Label Sounds” effect (in the “Analyze” menu).
The manual for this effect is here: file:///usr/local/share/audacity/help/manual/man/label_sounds.html

You will probably need settings similar to this:

Screenshot_2022-01-08_20-15-24.png

Well, that’s close, but there are a lot of false positives (about 16). Your suggested settings are pretty much what I tried except my Maximum Trailing Silence was set to 1.00 sec.

Thanks,
RoseM

To tune the settings for best accuracy:

  • Threshold: Just a bit above the level of the “silences”.
  • Minimum Silence Duration: A little bit less than the smallest silence that you want labelled.
  • Minimum label interval: A bit less than the shortest chapter.

That resulted in about 40 false positives.

I’m beginning to think that I can’t get there from here. I still think, for me, that I need a tool for setting the length of the chapter announcement.

Thanks,
RoseM

With what settings?

These are the settings:

LabelSource.jpg
Thanks,
RoseM

Surely all of the chapters are much longer than 3 seconds! Try increasing “Minimum label interval” to a bit less than the shortest chapter.

OK, here’s the latest attempt. Still false positives. When a false positive shows up it throws all remaining detections off. Only the first chapter break was detected - everything else was a false positive.

LabelSource1.jpg
I think, in my case, that Label Sounds is unable to do the job for me. Shrug!

Thanks,
RoseM

If the track contains silences over 2 seconds long that are not chapter marks, then of course they will be detected along with silences over 2 seconds long that are chapter marks. The best that can be done is to tweak the settings so that all of the chapter marks are selected, then manually delete the labels that occur in gaps that are not chapter marks.

I can manually find the chapter announcements faster than I can tweak the analysis tool then find the chapter announcements.

It’s becoming obvious that Audacity does not have a tool to help me find the chapter announcements.

Thanks for the help,
RoseM

Rose,

what I did when I was recording my vinyl and tapes was to Press Ctrl+M when the recording was in-between songs to drop a label at the current recording position.

Peter.

Hi Peter,

Thanks for the tip! I was not aware of Ctrl+M. I’ve only been using Audacity for a few months and I’m probably using about 1% of its capabilities.

Thanks,
RoseM

Rose, thank you for this topic.
I was looking for a way to determine the first “real sound” in a recorded track; with that in hand I reasoned that the leading “silnce” would be between that point at the start of track.
Untitled.png
These settings (different from yours!) seem to do the trick!
So thanks for the topic and hence my education.
Chris

Hi Chris,

Thanks for the tip. I tried your settings and it found all the chapter breaks (5), but it also had about 10 false-positives. I’ll play with it some more to see if I can tweak it a little.

Thanks,
Rose

Rose you are welcome, but i wasn’t trying to help you as much as to add more data to the thread.
As I understand it, you are trying to split one track into several sections (of your dad’s book), whereas I was trying to isolate what I think of as “the recorded sound” part of my track from the “quiet bit at the font” and the “quiet but at the end”.

I start off with a “chapter” that i have recorded by reading from a book. I have a bit at the front, “room noise”, where I settle my seat, slowly take a breath etc, before I start reading aloud. At the end of the chapter I pause for a few seconds and the recorder runs on, leaving a bit of “room noise” at the end.

I am new to Labels, and I was looking for (and - Thanks to You - finding!) the settings that worked for me to identify the audible noise that I want to retain, and the run-in and run-out that ultimately I want to drop.

I would think that if you used something close to my settings, your Dad would exclaim “That’s my chapter!”

Cheers
Chris