Hacker News Reader

Comment by jobigoud | Hacker News Reader

Regarding audio they say:

> Doubling the speed, in our experience, leaves individual words perfectly identifiable — but makes it just about impossible to follow the meaning.

Wow, sorry but this is very misinformed. A lot of people me included can listen to audio at 2x. I constantly listen to podcasts (so not a slow narration to begin with) at 2x in English which is not even my native language and I understand everything and enjoy it very much.

Blind users listen at 4x without loss of understanding.

You just need to train it.

View on HN

Seconded. After I saw and heard how blind people use screen readers (with experience), I started accelerating audio. You can train your brain to become better at understanding — I started at 1.25x and slowly worked my way up. I am now at 1.75x - 2x (plus Overcast's silence removal) and I have no problems, even though English isn't my native language.

Two requirements: headphones (you need much more processing power to deal with crappy speakers and poor audio in general) and lack of significant distractions (you can't afford to think about something else). Also, some apps have crappy audio acceleration algorithms. Overcast does it so well that I use it also for listening to audiobooks.

I also think the article is misinformed. You can speed read by regularly training your brain. Just don't expect miracles and don't believe the "experts" who try to sell you miracles.

Yes the speech rate algorithm is important. The application I use uses the Sonic algorithm by Bill Cox, which is actually optimized for rates over 2x.

https://github.com/waywardgeek/sonic

I think speed rate should be a standard control on every multimedia player, just like volume.

> The application I use

What application do you use for listening?

Well I haven't tested a lot of them so not sure if it's relevant but I use DoggCatcher as a player and another app called Presto that shims the Android audio layer to increase the speech rate transparently for the player app. It's Presto that uses the Sonic library.

Do you have a link for Presto? I can't find anything about it online.

probably https://play.google.com/store/apps/details?id=com.falconware...

Hmm, now you got me worried. I use a dedicated device for podcasts that is still running Android 2.x. Prestissimo seems to be the new name, but itself deprecated in Android 6.0. I'll have to look into that before that device dies.

I didn't know about Overcast. Volume normalization for listening in noisy places, smart speed with silence removal - I want those features!

Unfortunately Overcast is iOS only. Could anyone recommend an Android app similar to Overcast?

Pocket Casts [1] is the closest you can get on Android.

I use Podcast Addict [2], but it's a bit bloated to be frank.

[1]: https://play.google.com/store/apps/details?id=au.com.shiftyj...

[2]: https://play.google.com/store/apps/details?id=com.bambuna.po...

Pocket Casts is amazing. I use it on iOS and Android. It has silence removal and variable speed playback.

Well, before Overcast appeared I used to just manually process audiobook files with ffmpeg. The command line was:

`ffmpeg -threads 16 -i "$infile" -vn -filter:a "atempo=2.0" -c:a pcm_s16le "$outfile"`;

this produced a .wav file, which is what I fed into Audiobook Builder to produce an .m4b audiobook.

Overcast does a significantly better job, though, and without any extra hassle on my part.

It depends on the density of the information doesn't it? I would certainly agree that many, if not, most podcasts could be listened to at double speed or higher and still be understood. However if the podcast is about a detailed analysis of how string theory explains quantum gravity effects, then I highly doubt one could listen to it at double speed and retain much understanding.

I'm a physics student who has listened to tons of physics course videos at 2x or sometimes 2.5x. I pause when an equation is drawn on the board, but the speech is fine.

The inverse holds that the density is sometimes not high enough. I had trouble following videos on the mill cpu at normal speed, but at 1.25/1.5 my mind was forced to pay more attention

This part made me question the entire article. Just about any audiobook or podcast under 2x is too slow for me and most videos (including tv shows and movies) I watch at 1.5x comfortably. I'm also much faster at reading today than I was a few years ago. There is definitely a way to get much faster (while still being realistic).

Yeah. For a long time I thought I hated nearly all video-based entertainment. Then Youtube got that speedup feature. Now that I can watch things at 1.7x, it's much more interesting.

One of the biggest limits is the processing I think. Good algorithms try not to omit anything, but those also begin to blur vowels in to consonants. Fine-tuning of the window size helps (not sure if this was on the FFT or resynthesis part).

Close to the speed limit for understanding two things are difficult:

- personal names (especially of companies) and aberrations.

- a change of the speaker to someone who has a slightly higher speed or worse audio quality

Unavoidable and super annoying:

how back in the "real world" everything appears to be in slow-motion. It takes ages for someone to just finish a single word...

I second this. English is my second language and if I listen to an audiobook or watch an informational video in normal speed, I might as well not do it at all... because my brain starts thinking of something else if it is too slow (i.e. normal speed) and I lose track of it.

Related: There's a plugin for chrome that lets you speed up all html5 videos. I usually watch youtube at at least 1.5x. The plugin is better than the youtube settings because it lets you set the speed on small increments.

https://chrome.google.com/webstore/detail/video-speed-contro...

Plugin schmugin. Try my bookmarklet:

javascript:void%20function(){document.querySelector(%22video%22).playbackRate=parseFloat(prompt(%22Set%20the%20playback%20rate%22))}();

Works on all HTML5 video that's not inside an iframe or otherwise inaccessible to the page-level DOM.

Only tested on FF for Windows and Linux, but it's just vanilla JS.

Perhaps this depends on the speed of your brain? You know, different people score differently on IQ tests. For example, some score 140 while others score 70. I read (in the book IQ and the Wealth of Nations by Richard Lynn and Tatu Vanhanen, I think) that what an IQ test basically does, is measuring the speed of your brain.

So I guess intelligent people will be able to understand everything on 2x while dumb ones may not.

I was on a debate team in college, and understanding people talking at 200 to 300 wpm is a critical skill. Wired ran an interesting article on this phenomenon a few years back: http://www.wired.com/2012/01/ff_debateteam/

All these responses about subjective, andecodtal experiences are interesting, but is there any data or any research to support it?

I don't have research about 2x audio to cite but I think there are datapoints that the authors didn't consider when they wrote "Doubling the speed [...] makes it just about impossible to follow the meaning."

Consider that most of us can read text at a normal unhurried speed of 200wpm to 300wpm. We also hear this "inner voice" as we read the written words. This subvocalization is therefore "sounding out" the words at ~300wpm in our head.

However, many people speak out loud at only ~100wpm. For many of us, accelerating speech from 1x to 3x is simply making the speaker sound out the words at 300wpm. Since that's the same as our subvocalization wpm, the meaning is not impossible to follow. Most humans can't move their mouths fast enough to talk at 300wpm -- but with digital technology -- they don't have to.

Spoken recordings at 100wpm is mind-numbingly slow and it would just make my mind wander.

I think the authors should have surveyed a hundred youtube users that always take advantage of the 2x speed option. They would have been surprised to learn that people can follow the meaning of the words at 200wpm very easily.

> there are datapoints that the authors didn't consider

In fairness, 1) we don't know what the authors considered; they apparently have done a lot of research on the subject, and 2) what you list aren't data

>we don't know what the authors considered; they apparently have done a lot of research on the subject

That's true we don't know what they considered but a cursory search doesn't find any primary research on 2x audio acceleration performed by either Rebecca Treiman or Jeffrey Zacks. As for reading text, it looks like RT's commentary on speedreading is based on meta-research[1] and not primary lab research. (It's also possible they published primary research in an academic journal that google doesn't index.) It looks like their comment about speech acceleration is anecdotal instead of being based on rigorous research.

>what you list aren't data

Why not? If an audiofile has speech content, one can count the total words, divide by total minutes, and hence, calculate a words-per-minute. Why is wpm not data? The average reading speed of adults is also well-documented: https://www.google.com/search?q=average+reading+speed+of+adu...

In any case, HN readers can try it themselves on a bunch of youtube videos.

For example, the Scarlet Letter novel: https://www.youtube.com/watch?v=KnfaAZu0fNE

Another example from a news show: https://www.youtube.com/watch?v=ZTSr07u8gow

Apple presentation with Tim Cook: https://www.youtube.com/watch?v=8zbAFSJCPSI&feature=youtu.be...

Change the playback option to 2x on those videos. Can it really be impossible to understand the meaning as the authors assert?!? For listeners whose native language is not English, perhaps they require the slower 1x playback speed. For most others, I'm very skeptical that 2x speed is impossible to extract meaning.

[1]http://www.psychologicalscience.org/index.php/publications/s...

Holy crap, Tim Cook at 3x is still too damn slow.

Completely agree, you just have to work your way up. Made a custom chrome extension and often watch videos at 4x speed (depending on the speech rate of the actual narrator).

Reply on news.ycombinator.com