Learning through audiobooks

Post Reply New Topic

Hello,
I have recently discovered it is very simple to automatically make audio files of books,
and also download lectures as mp3s from youtube.

This provides a great opportunity to learn while doing other things, like wandering around or cycling.

At the moment I am learning about politics like anarchism, and a new language.

Has anyone any tips or thoughts on this mode of learning?
I have noticed it is a bit easy to tune out, so I wonder if people have thought about this before.

^That's my problem with audiobooks. I just tune out. Never managed to overcome it so i go back to reading.

I find paraphrasing works when learning something through audio.

Summarising what the narater said in your own words.

It keeps my attention on the topic at hand

I go through spurts where I listen to a lot of podcasts in my target languages. I’m less likely to tune them out if they’re on topics I’m interested in. In my experience, you need to be at least an early intermediate language learner for content to be interesting, but it doesn’t take too long to get there.

What I like about it:
1. Hands-free
2. Gives my poor old eyes a rest from close-up reading.
3. Gives my neck a rest from staring at a screen or page.
4. May in some circumstances make it easier to take in the information.

What I dislike about it:
1. Goes at its own pace, which might not be my preferred pace, harder to skim and navigate to bits of interest than text is.
2. The speaker's intonation might not be suitable for me.
3. Depends on machines at point of use.
4. May depend on using a particular website every time I want to listen (tracking, ads, possibility of media vanishing, Internet outages etc.), unless I've figured out how to capture the content as a personal copy.

In fact I've been looking for a good way of converting speech to text. This has given me excellent results in terms of accuracy and simplicity:

https://mvsep.com/en

Upload your audio file, choose Whisper as the separation type, click Separate, and wait till it gives you the result. But it has a strange tendency to occasionally repeat a phrase. Also, unless you pay them, the wait time can be very long and it'll only process 10 minutes' worth of audio at a time, so converting an entire audiobook is pretty much out of the question.

Suggestions for a more practicable method would be very welcome. Can Cortana do it properly if I tell it I want to dictate a letter?

As for text to speech, I think there's something in Word that may do that. But I've only ever used a tiny old utility called SayIt, which sounded like Stephen Hawking on a bad day, so it was of limited use. I tried the Word thing once, but there was something limiting about it, can't remember what.

I haven't tried audiobooks but I often watch documentaries and educational videos on YouTube. I think the videos are okay, fairly informative, and I find them less demanding than reading. But I do remember things I have read on a page (in a book rather than online) better. I remember seeing the words if that makes sense. But I suppose for casual learning audiobooks might be okay, for me.

You can try google text to speech, which is ok for free, not amazing like whisper, but better than word. You may need to write a python script or use some web service, probably someone wrote a wrapper somewhere.

By all means use anything that works for you. I can't stand listening to the incorrect cadence and pronunciation of words by synthetic voices, so it doesn't work for me.

I too dislike those synthetic voices, though I might be able to get over that if they didn't also give me practical problems in focussing on what they were saying. There's something about those faults in the delivery that throws my focus every time. That only happens with a human speaker when they're reading aloud and aren't very good at it.

I've read that they've made a lot of progress with cadence, pronunciation etc., but I've yet to hear the results.

Sadly, I avoid Google because of the tracking. So I looked around the Web again.

This website gives mediocre results:
https://www.cross-plus-a.com/balabolka.htm
Even with the extra voices from here:
https://rhvoice.org/en-voices/
it's still mediocre.

I thought I was onto something when I saw how pleased some people were with ClipChamp:
https://redlib.privacyredirect.com/r/ga ... nvertible/
https://clipchamp.com/en/

But Microsoft took it over and seem to have made it so that you can't use it without Edge

Then I saw this, which is supposed to use ClipChamp's voices:
https://crikk.com/text-to-speech/

But in spite of the rave reviews, IMO those voices are only marginally better than Balabolka. Glad I didn't obey MS and get Edge, if that's the best ClipChamp is capable of.

The search continues. I doubt any of the things I've tried so far are actually AI. You may be right that I'd need to do something very technical to get it for free. GitHub may have the answer, but that place is Greek to me. It's funny because people see me as a very technical person.