frenkeld Diglot Senior Member United States Joined 6859 days ago 2042 posts - 2719 votes ![](/images/pokal.2.jpg) ![](/images/pokal.2.jpg) ![](/images/pokal.2.jpg) ![](/images/pokal.2.jpg) Speaks: Russian*, English Studies: German
| Message 1 of 44 28 November 2006 at 12:41pm | IP Logged |
This may already be known to all, but just in case...
It turns out that the current 1.3.2 beta version of Audacity free audio editing software not only has a "Truncate silence" feature, but also a rudimentary, yet sufficient, scripting-like capability, which can be found in this version of Audacity under the "File" menu.
What's needed is "File -> Edit Chains" and "File -> Apply Chain". In the former, one can look at the "Clean Speech" and the "MP3 Conversion" chains for two examples of what to do, and then one can easily write one's own chain (my first cut was just "Stereo to mono" -> "Truncate silence" -> "Export mp3"), which can then be applied to a set of files by the "Apply Chain" command.
So, one can take a whole set of Assimil lesson files and quickly reduce the gaps in all of them at once.
One would probably want to edit the first six lessons by hand anyway, to get rid of the two copies of the dialogs, maybe speed up the tempo, as well as knock out the gaps, but the rest of the lessons should be a breeze, if you don't mind keeping the exercises with the dialogs.
Edited by frenkeld on 28 November 2006 at 12:42pm
3 persons have voted this message useful
|
Farley Triglot Senior Member United States Joined 7008 days ago 681 posts - 739 votes ![](/images/pokal.2.jpg) ![](/images/pokal.2.jpg) ![](/images/pokal.2.jpg) 1 sounds Speaks: English*, GermanB1, French Studies: Spanish
| Message 2 of 44 28 November 2006 at 1:21pm | IP Logged |
Thanks for the tip. I just gave it a try. What a time-saver!
1 person has voted this message useful
|
Kveldulv Senior Member Italy Joined 6869 days ago 222 posts - 244 votes ![](/images/pokal.2.jpg) ![](/images/pokal.2.jpg) 1 sounds Speaks: Italian*
| Message 3 of 44 28 November 2006 at 2:37pm | IP Logged |
Good advice, frenkeld. I was tired of working on a single file at a time.
1 person has voted this message useful
|
Farley Triglot Senior Member United States Joined 7008 days ago 681 posts - 739 votes ![](/images/pokal.2.jpg) ![](/images/pokal.2.jpg) ![](/images/pokal.2.jpg) 1 sounds Speaks: English*, GermanB1, French Studies: Spanish
| Message 4 of 44 28 November 2006 at 4:33pm | IP Logged |
FYI,
I experimented with the settings and found that defaults parameters for Truncate Silence: Db=-40db Duration=1000; will reduce a Pimsleur file down to under 15 minutes.
I had to set the Duration=500 to get ride of the pauses in Assimil files.
1 person has voted this message useful
|
frenkeld Diglot Senior Member United States Joined 6859 days ago 2042 posts - 2719 votes ![](/images/pokal.2.jpg) ![](/images/pokal.2.jpg) ![](/images/pokal.2.jpg) ![](/images/pokal.2.jpg) Speaks: Russian*, English Studies: German
| Message 5 of 44 28 November 2006 at 5:10pm | IP Logged |
I used a -35db threshold for "silence" on "German with Ease" ripped with EAC ("Exact Audio Copy").
The way I came up with it was as follows. I pulled out the floating toolbar with level meters and stretched it across the screen to see the numbers well. Then I selected a silent interval and "played" the selection once. The meter showed the maximum level attained. There are intervals which are really silent, but which look just a bit wiggly on the waveform, and they, as I found, go above -40db, but don't reach -35db. So, I used -35db to cut them out and that worked well.
Of course, if you go too high, you'll start cutting out speech as "silence", so one has to be careful.
P.S. Once "silence" is correctly determined for the recording, I would expect the time parameter should just be the maximum allowed silent interval in milliseconts.
My guess is that noisy recordings (say, those made from tapes) may need to have the noise cleaned up first for the software to figure out the silent intervals correctly. This version of Audacity seems to have some noise clean-up functions too, but I haven't had the need to try them so far. The "Clean Speech" chain incorporates noise removal.
Edited by frenkeld on 28 November 2006 at 5:51pm
1 person has voted this message useful
|
Sir Nigel Senior Member United States Joined 7020 days ago 1126 posts - 1102 votes ![](/images/pokal.2.jpg) ![](/images/pokal.2.jpg) ![](/images/pokal.2.jpg) ![](/images/pokal.2.jpg) 2 sounds
| Message 6 of 44 28 November 2006 at 6:51pm | IP Logged |
Would it also work to use a noise gate first?
In other words the silent parts (slightly noisy or cleaned up) would become digital silence and then the truncate noise feature could be set lower, perhaps to -70. This seems more accurate and wouldn't risk removing speech as I assume the noise truncate feature doesn't have a variable attack speed.
1 person has voted this message useful
|
frenkeld Diglot Senior Member United States Joined 6859 days ago 2042 posts - 2719 votes ![](/images/pokal.2.jpg) ![](/images/pokal.2.jpg) ![](/images/pokal.2.jpg) ![](/images/pokal.2.jpg) Speaks: Russian*, English Studies: German
| Message 7 of 44 29 November 2006 at 10:54am | IP Logged |
I've experimented on "New French with Ease" by hand, and it turns out that if you apply "Truncate Silence" twice in a row (with the same parameters), it will do additional gap reduction on the second pass, but the changes on further passes are less likely.
So, I modified the original 3-step chain I wrote for Audacity by having it now do two "Truncate silence" operations in a row, and this seems to work pretty well.
"New French with Ease" must've been mastered digitally, so -40 db seems to be enough to cut down all the silient intervals to specified size, but only if "Truncate silence" is used twice in a row.
So, my new chain is: "Stereo to mono" -> "Truncate silence" -> "Truncate silence" -> "Export mp3"
Edited by frenkeld on 29 November 2006 at 12:26pm
1 person has voted this message useful
|
Vinnie Groupie England Joined 6440 days ago 65 posts - 66 votes ![](/images/pokal.2.jpg) Speaks: English*
| Message 8 of 44 20 January 2007 at 8:34am | IP Logged |
I would like to do this to my Italian assimil. I know you have explained it here but it doesnt seem to work for me. I copied lesson 2 and then tried playing it in the editing thing but it just makes a strange noise for a second. It wont play and i have not got a clue how to use it.
Could you direct me somehow?
Thanks
1 person has voted this message useful
|