This is a great site!
I think about using a few audio books from this site for creating some books in the audio-text synchronized format with the software I am developing.
http://www.interactiveselfstudy.com/
I am creating a program for listening audio books.
I've got to say that's a fascinating concept you're working with KD. Thanks for showing us. I sure hope it works out well for you; turning good ideas into cash flow is an art unto itself
-Chip
Retired to Colorado
The man who does not read good books has no advantage over the man who cannot read them.
~Mark Twain
Retired to Colorado
The man who does not read good books has no advantage over the man who cannot read them.
~Mark Twain
-
- LibriVox Admin Team
- Posts: 7972
- Joined: September 26th, 2005, 4:14 am
- Location: Montreal, QC
- Contact:
hi KD cool project. you might want to check out a LibriVox spin off project (not sure where it's at), called revoxer:
http://revoxer.sourceforge.net/
http://revoxer.sourceforge.net/
Hi Konstantin,
It's great that we have more than one project that
tries to make LibriVox content more accessible!
Could you tell us a little about the format of your
files? I know that there are existing standarts for
synchronized audio/text. Do you use any of them?
-umut
It's great that we have more than one project that
tries to make LibriVox content more accessible!
Could you tell us a little about the format of your
files? I know that there are existing standarts for
synchronized audio/text. Do you use any of them?
-umut
Actually I found the revoxer first and from there I came to the Librivox website.hugh wrote:hi KD cool project. you might want to check out a LibriVox spin off project (not sure where it's at), called revoxer:
http://revoxer.sourceforge.net/
The format is quite simple in order to save time for development, because it was used as a demo and actually still is remaining on the demo level. The program can play only wav files and the interface is still on a very basic level.tshirt wrote:Hi Konstantin,
It's great that we have more than one project that
tries to make LibriVox content more accessible!
Could you tell us a little about the format of your
files? I know that there are existing standarts for
synchronized audio/text. Do you use any of them?
-umut
I've just put one of the stories on the web site (only it is only third synchronized and I've not found yet translations other than in Russian).
If you don't want to download it with the wav file just use http://www.interactiveselfstudy.com/downloads/TamingTheBicycleNoAudio.zip
You can see the format there. File book.txt contains text of the book, files book_t1, book_t2 ? are containing translations to other languages and file book_snd.txt contains the information how the sound corresponding with the text. In the other books this file is inside an exe file. All files are Unicode. Each sample is described by two lines: first line contains an index of the paragraph and indexes of the starting letter and last letter inside the paragraph, the second line contains or translation or sound period. In the beginning I edited those files manually. Later I wrote a few programs to make the synchronization process faster. For sound synchronization it is based on the main program, parts of the text and the corresponding sound samples are selected by mouse clicks and it looks like this.
I think it takes about 5-7 minutes to synchronize 1 minute of audio.
It uses voice activity detection by analyzing a wav file (actually very simple algorithm).
Some other small features are hardcoded for each book separately: go to chapters, illustration positions, language menu of available translations just because of lazyness.
If someone wants to try to synchronize some book, I can put these tools on the web site and describe how to use it.
Wow, this sounds great. Do you need user input for this algorithm?KD wrote: I think it takes about 5-7 minutes to synchronize 1 minute of audio.
It uses voice activity detection by analyzing a wav file (actually very
simple algorithm).
I am guessing that 5-7 minutes is the computation time.
ReVoxer design uses human input for synchronization. Have you ever
tried such an approach? With an intuitive user interface, that could
save a lot of time for synchronization, i.e. about the same time for
annotation and listening a recording. We are hoping that as more people
annotate the audio the synchronization becomes better.
Again, looks like you did a great job there. Thanks for sharing it with us.
Do you think you can port it to run as a java applet?
Hello,
For full automation of this process was necessary to use some speech recognition - it was too much work and narrowed to the particular number of languages, may be later I will write something more productive.
I put this ?Synchronizing package? onto the web site http://www.interactiveselfstudy.com/downloads/TamingTheBicycle_map.zip so you can try it.
It is not very user friendly but still is simple enough. I tried to write some kind of instruction.
- Copy the wav file also into the ?map? subdirectory,
- Run SoundMapper.exe, give Noise level some number between 100-300. It should produce two files, book.txt and book_snd.txt in the subdirectory.
- Run InterBooksSynchSound.exe, it opens a window in the bottom, showing time stamps of the sound ranges.
- Run InterBooksSynch.exe.exe, it opens a window on the top with the text of the books.
- Play audio in the bottom window using right mouse button for viewing what time stamps forming complete phrases.
- Double click in the upper window with the right mouse button into the first symbol of the phrase and into its last symbol, it should change its colour into brown or magenta.
- Double click in the lower window with the left mouse button into the first time line of the phrase and into its last or the same line, it should change its colour into blue or red.
- After repeating this procedure a few times in the file menu of the upper window choose ?Reload Sound Synchro? option and check how it was synchronized by playing the audio in the upper window and you can see the synchronized time stamps also in the book_snd.txt file in the main directory.
Yes, it requires user input, by double clicking into the text. But this process is fast enough, almost like playing the Minesweeper.tshirt wrote: Do you need user input for this algorithm?
I am guessing that 5-7 minutes is the computation time.
For full automation of this process was necessary to use some speech recognition - it was too much work and narrowed to the particular number of languages, may be later I will write something more productive.
No, it is written on C++.tshirt wrote: Do you think you can port it to run as a java applet?
I put this ?Synchronizing package? onto the web site http://www.interactiveselfstudy.com/downloads/TamingTheBicycle_map.zip so you can try it.
It is not very user friendly but still is simple enough. I tried to write some kind of instruction.
- Copy the wav file also into the ?map? subdirectory,
- Run SoundMapper.exe, give Noise level some number between 100-300. It should produce two files, book.txt and book_snd.txt in the subdirectory.
- Run InterBooksSynchSound.exe, it opens a window in the bottom, showing time stamps of the sound ranges.
- Run InterBooksSynch.exe.exe, it opens a window on the top with the text of the books.
- Play audio in the bottom window using right mouse button for viewing what time stamps forming complete phrases.
- Double click in the upper window with the right mouse button into the first symbol of the phrase and into its last symbol, it should change its colour into brown or magenta.
- Double click in the lower window with the left mouse button into the first time line of the phrase and into its last or the same line, it should change its colour into blue or red.
- After repeating this procedure a few times in the file menu of the upper window choose ?Reload Sound Synchro? option and check how it was synchronized by playing the audio in the upper window and you can see the synchronized time stamps also in the book_snd.txt file in the main directory.
How do you make the exe?
What's this little box thingy for? Oh! [color=red]C[/color][color=orange]O[/color][color=yellow]L[/color][color=blue]O[/color][color=indigo]R[/color]
I added general exe file into the zip file, but now for each book exe file is created with some hardcoded features: "Go to" and "Translation" menus and illustration positions. I did it in order to save time for development. Probably, when number of such books grows I will make it more general -with one exe file for any book. If you want to make such book yourself I can build such exe file for you.ab2525 wrote:How do you make the exe?
cool... can you do that for me?
What's this little box thingy for? Oh! [color=red]C[/color][color=orange]O[/color][color=yellow]L[/color][color=blue]O[/color][color=indigo]R[/color]
Sure, only I need more information - text, translations, illustrations.ab2525 wrote:cool... can you do that for me?
My E-mail isslb@canada.com.
what if there are no translations or illus.?
What's this little box thingy for? Oh! [color=red]C[/color][color=orange]O[/color][color=yellow]L[/color][color=blue]O[/color][color=indigo]R[/color]
No problem. It is optional but not required. Do you have the text and audiofiles?ab2525 wrote:what if there are no translations or illus.?