I just discovered, at least for myself, what I think is an innovative approach to learning from CD's and cassettes. I've always been plagued by the fact that the normal speed of Irish-language CD's and cassettes are much too fast for a beginner. I think it's the singular reason why most adults never learn how to speak Irish despite years of lessons and countless hours of listening to audio material.

As a former piano teacher I came to the realization that students, and they were usually adults, who tried to learn quickly never learned at all. It was the student who practiced slowly, who took the time to grasp the meaning of the notes and phrases that learned not only the quickest, but most successfully.

So, I'm delighted to have found a way to learn Irish slowly, at least with audio material. Today I took a track from Learning Irish and, using Audacity, http://audacity.sourceforge.net/, I was able to slow down the speed of the recording (Effect/change tempo) without changing the pitch of the recording, which is the usual result when trying to slow down an audio file. That is, I now have a digital recording of Chapter 7 at a speed and the original speaker's pitch that I can comprehend. I think that I'll finally make progress because I can now learn slowly enough to comprehend what the speakers are saying.

I want to thank Aaron for introducing me to Audacity's software - go raibh míle maith agat a Aaron!

I would also like to get input (no pun intended) from other learners who might have tried such an approach or would be interested in trying the approach out themselves.

Hey, you're welcome, Mac Léinn!

I'm glad to see that people are making use of Audacity, I've used it quite a bit to alter audio materials. Slowing the tracks down is great. I like to listen to the slow track and then the regular-speed track, so I have an idea of what the regular-speed track is going to say. Sometimes I put the regular track first, and the slow version after, so I can figure out what they said.

I also use Audacity to edit tracks, often to remove the English from them so it's all i nGaeilge or in nWhatever I'm studying.

I've also used it to record audio off the net so I could study languages in actual use. Not learners material, but audio broadcast on online radio or "tv", and then I'd make a track (usually slowed down) and listen to it over and over to try and understand.

If anyone needs help with Audacity let me know, I'd be happy to help.

Haigh a Aaon,

Yeah, the uses for Audacity are endless. Is pianodóir mé, and I've never been able to successfully transcribe live recordings, but now with Audacity, I should be able to slow the pieces down sufficiently to transcribe songs now without losing the proper pitch.

Question: Using Ó Siadhail's Learning Irish as an example, I notice that after slowing the tempo, there's a slight "warbling" or shakiness to the recordings. I can still hear the speakers clearly, but I would like to know if there's additional post processing that I can do with Audacity. I'm somewhat familiar with sound waves, having done my master's in Acoustics, and know a little bit about filtering to get myself in trouble. But I'd like to know if you've experienced this same phenomena and what, if anything can be done?

I had gone to the extent of transferring my Learning Irish cassette to CD's using Audacity some months ago, but I think I'll buy another copy of the book, this time with the CD's to avoid the noise that has entered my recordings from the cassette recorder used when transferring the cassettes to CD's.

As you've mentioned above, using Audacity to copy stuff off the net is great also. Just curious as to why you put the slow track after the regular-speed track. My plan is to put the slow track first so that I can understand what is being said in the subsequent regular-speed track. I found this technique of slow first, then regular speed after very effective in my own piano playing and also in teaching piano students. Again, just curious.

Also, it's very "audacious" of you to offer your help to us - just a play on words there if you don't mind - it took me some while to figure out why they called the software "Audacity" - duh!

quote:

there's a slight "warbling" or shakiness to the recordings. I can still hear the speakers clearly, but I would like to know if there's additional post processing that I can do with Audacity.

Do you mean the sort of high-pitched reverb sound? That can usually be avoided by using the speed change and pitch change together instead of the tempo change. That's basically what the tempo change is - slow down the speed, then raise the pitch back to normal. The cause of the problem (as far as I can tell) is some of the high frequencies don't go through the pitch change algorithm very well. I've noticed deeper voices don't have the warbling problem as much as the higher-pitched voices do.

Try this on the original track:

1. Drop the speed down to about -50%. It will sound like it is in super slow-motion. This gives the computer very low frequencies to process instead of high frequencies.

2. Adjust the pitch up about 30% percent.

3. Raise the speed back up by about 60% percent.

Try different percentages to get what you're looking for. If you want to do some math, you can get the pitch back exactly, but I usually just try to get it in the ballpark. If you're slowing down a fairly high voice, or still hear some warble, slowing the track down farther should help.

quote:

to avoid the noise that has entered my recordings from the cassette recorder

Did you try using the noise reduction? I've had various levels of success with it.

quote:

Just curious as to why you put the slow track after the regular-speed track.

I actually usually put the slow one first, but having the slow one second is sort of like taking a test and then looking at the answer key. I try to understand it fast, then listen to it slow to double check. I usually listen to audio using my mp3 player, flipping back and forth with the skip buttons, so the order is totally flexible. I'll listen two or three times slow, then try it fast. If I was to burn it on CD, I'd put the slow first.

Do you mean the sort of high-pitched reverb sound?

Yes, exactly.

The cause of the problem (as far as I can tell) is some of the high frequencies don't go through the pitch change algorithm very well.

I work in the telecommunications industry with many very gifted electrical engineers, one who is somewhat familiar with the algorithms in this field of sound-processing that were invented by our previous parent company. He tells me that the math behind these algorithms is quite intense - Hilbert transforms I think. I'll ask him about the high-frequency issue and see what he thinks. He works on truly high-frequency stuff, to the tune of 1 giga hertz! I'm more familiar with frequencies in the audible range, so even though he and I are on "different wavelengths" maybe we can gain a better understanding of the problem so that we can tweak the controls to our advantage.

But I'll start fiddling with the controls per your advice above. And I'll definitely try doing the math knowing that if I really screw things up I can just start over again! I'd like to keep the same pitch if I can so I'll probably write out some ultra-simple algorithm (aka equation) that follows your suggestions above.

Go raibh míle maith agat as do chuidiú.

Mac Léinn Fuaimíochta

(Message edited by mac_léinn on February 27, 2007)

I'll ask him about the high-frequency issue and see what he thinks.

Great, I'd be interested in knowing his thoughts on this.

Let me know how things go with slowing down the Gaeilge tracks and getting rid of the warble.

Haigh a Aaron,

I spoke with my elecrical-engineer colleague this morning who's familiar with the warble effect. The first thing he did was play back one of his voice mails and he slowed it down. His purpose in doing this was to show me that some people (our previous parent company's engineers) have figured out the problem, which was clear to me since the slowed down message didn't exhibit any warbling.

We got to talking about Fast-Fourier Transforms and how the slowed down tempo is achieved by breaking the complete sound recording into smaller sections, perform some high-falutin' numerical filtering, then re-stitching the transformed sounds bite back together. My EE colleague indicates that it's that re-stitching of the sound bites that cause the warbling.

Well, that's what I've learned so far. Unfortunately we have to work here at work, so he and I will discuss this in more detail and I'll keep you informed.

Mac Léinn Trasfhoirm Fourier