MegaVoice Audio Processing

Overview of MegaVoices
Ambassador
Scribe
Audio Processing
Clamshells
Instructions
Tamper Resistance
Alternative Power
Bible Book Introductions
Projects

Audio processing for the MegaVoice is a multi-step process. Here's an overview:

1. Source recordings are made, frequently by missionaries on the field, or by nationals. The quality of these source recordings may limit the compression ratio used in the final encoding process by MegaVoice, so it's important that these recordings be the highest quality possible--even though they will later be sampled down to roughly "telephone quality" audio.
2. Recordings are converted to WAV files by "chapter" or other desired units.
3. WAV files are filtered, preparing their audio qualities for high compression ratios.
4. WAV files are split into small files and named hierarchically, corresponding to the three levels of forward/back buttons on the MegaVoice units.
5. After final processing of WAV files, these are sent to MegaVoice for compilation.
6. The compiled files are ready to be loaded on the Ambassador units.
   1. Ambassadors can be ordered from MegaVoice pre-loaded by project number.
   2. The compiled project files can be returned from MegaVoice, ready to load on the Scribe from a computer, then loaded onto Ambassador units.

Now for the details.

NOTE: the procedures below should be considered as a DRAFT version ("alpha" or "beta"), to be refined through testing and usage.

1. Sound Recordings

Made in the field....

2. Recordings Converted to WAV chapter files

Rip Audio Files

 

Using Creative Mediasource player, rip files from audio CD

Album – Luke (book or major division on MegaVoice)

Artist – Cabecar

Genre – Cabecar

In Tools > Settings > Audio CD/Rip tab > Rip Format: Wav, Mono, 44100 kHz, 16 Bits Per Sample

 

Recordings are placed in: C:\My Music\Cabecar\Album directory (where Album name is specified above)

 

Resulting files are BIG (about 5MB per minute of recording), so have lots of hard disk space available.

 

Rename Files

 

Using the ReName file utility

l        can be downloaded from

l        for ease of use, put a copy in each directory where you want to use it

l        use F2 to go to Expert mode

l        locate an “Album” (book) directory, and rename files in following format:

 

            00000-Luk001-nfC001.wav

 

00000	Luk	001	nf	C001
MegaVoice  project number—determined by MegaVoice—use zeros as placeholders	3-letter book abbreviation   use Paratext abbrs.	Book sequential number for purpose of this MegaVoice project—determines order of major divisions on the player	2-letter language code   cb = Cabecar	Chapter sequential numbering—determines order tracks are played.  Note that if you have an intro section, numbering will not correspond to “chapter numbers” in Bible

 

The above format is actually the format for chapter titles in the final processed files, but it in the intermediate stage it works well to label the entire chapters this way.

 

3. Filtering WAV files

Using NCH Swift Sound WavePad

 

I am using WavePad v 3.00 from NCH Swift Sound (http://nch.com.au/wavepad/index.html), a program that sells for around $50-75.  I already had this program on my computer, so I thought I would see if it has the capabilities required for to prepare files for the MegaVoice devices.

 

I am outlining in detail the procedures I am using to prepare sound files for MegaVoice in order to confirm that they are adequate, for myself and other team members here to follow, and even to offer as a viable alternative to Adobe Audition if the results are in fact adequate.  Potential benefits include:

l        More cost efficient (15% of the cost of Audition)

l        Batch processing capabilities

�       batch script can be saved for reuse

�       reduces potential for human error

�       decreases time required

�       increases consistency

�       more efficient use of disk space (no intermediate files required)

 

(I will also note that Audacity will have Batch processing available in its next release (1.3), and even has a built in Speech processing script which appears to have similar parameters to the MegaVoice procedures.  This might provide another option in the future, but the current Beta release 1.3.0b is very unstable and useless at this point.)

 

Source Files:

 

Our source material was available on audio CD.  This material was ripped using Creative MediaSource v 3.20.79 to a file/directory structure, saving the output as PCM wav 16-bit mono files sampled at 44.1 kHz.

 

Source files are saved to C:\My Music\Cabecar\Luke

File Names: 00000-Luk001-nfC001.wav for chapter 1

Output files (generated below) are placed in C:\My Music\Cabecar\Luke\silenced

 

Batch Processing in WavePad

 

Batch processing in WavePad is set up as a 3-step process in the interface to specify: 1) files to to be processed, 2) batch commands, 3) output format and directory.

 

The Batch processor can be accessed in at least 3 ways: an icon on the icon bar, a command link on the left-hand command bar, and under the Tools menu (see screen shot).  Any of these goes directly to Step 1.

 

 

 

Batch Converter Step 1:

 

Use the Add Files button (the Add Folder button appears to have a bug in it and does not end up processing any files—see discussion at http://nch.invisionzone.com/index.php?showtopic=2404).

 

Click Next.
 

 

Batch Converter Step 2:

 

Browse up to C:\My Music and locate Speech8:

<<NOTE: later save in subdirectory with files—for convenience and so can reproduce same settings>>

 

Initially, the commands shown must be added using the Add button.  The settings shown are set by double-clicking the Command or clicking its Settings button.  However, see below regarding the Equalizer settings.

 

The Commands can be re-ordered by dragging and dropping the Commands in the list.

 

After the commands have been set up, they can be saved for reuse by clicking Save As and specifying an appropriate name.  Note the location where this file is saved for future reference.  These files can also be transferred to others for their use.  In the future, the saved set of commands is loaded by opening the saved batch file.

 

Click Next when Commands are properly entered.

 

Notes:

 

I have used Auto Trim to take off leading and trailing silences in the source files so that appropriate silences could then be added on using MegaVoice standards.  I also put Noise Reduction ahead of these steps so that the Auto Trim command would function properly.

 

I have inserted Amplify in 2 places since the Normalize doesn't seem to be working well.  Depending on the level of the source file, 90% x 90% may work well, or 80% x 80%.
Equalizer Settings

 

The Equalizer settings appear to be saved on a program-level basis, and are NOT retained by the Batch Script File.  Therefore, these settings most be opened and set the first time through, and should be verified prior to re-running any batch procedure.  Opening the Equalizer Settings (by double-clicking the Equalizer Command line, or by clicking the Settings button) reveals:

 

 

The HighPass setting is used with the Logarithmic Mode.  While the settings are displayed graphically, the actual parameters must also be reviewed.  These are revealed by clicking on the Settings button:

 

 

The Pass Frequency determines the top bend point, and subtracting the Slope Length determines the bottom bend point.  The Amplitude determines the percentage of the filter below the bottom bend point.  Thus the example above runs at 33% from 0 up to 100 Hz, then slopes up to 2000 Hz where it reaches 100%. 

 

I have tried a higher pass frequency using my source materials (such as 4000 following MegaVoice's example in Step 6, Figure 5), but the results did not seem as clear audibly.
 

 

Batch Converter Step 3:

 

 

Here the format is chosen as .wav and the output folder is specified.  In addition, the Format Options button reveals the following options for wav files:

 

 

The Format and Attributes must be chosen as PCM and 8.00 kHz, 16 Bit, Mono, and these settings can be saved as “MegaVoice” for future use using the Save Ass button (so in the future, this option will show up in the Name list).

 

Processing time seems quite reasonable, and results appear to be good.  Batch processing of the book of  Genesis took 8 minutes on a 5(?) year old 1.4 GHz AMD Athlon machine with 512 MB RAM.

 

4. WAV files split and named hierarchically

The open source audio editing program Audacity (available at http://audacity.sourceforge.net/) provides the best capabilities for splitting the filtered WAV files into the smaller segments that correspond to each press of the forward/back button keys.  Use of version 1.2.x has proven stable and is described below.  (Early Beta versions proved unstable, but also promise the possibility of batch processing in the future, which could potentially replace use of WavePad above.)

There are two methods for splitting speech files, while music files are split automatically:

1. Speech files can be manually tagged for splitting at actual paragraph breaks using a transcript of the recording.  Expected processing time is 3 times actual play time.
2. Speech files can be tagged for splitting using a semi-batch process by identifying natural breaks in the speech patterns in the recording.  Expected processing time is less than actual play time.
3. Music files can be split automatically based on approximately 20 second segments.

The procedures below are for method #2, but could be adapted for method #1.

Overview of processing:

1. load chapter file into Audacity
2. use Analyze > Silence Finder to identify potential breaks in the chapter.
   1. different settings may be needed in different parts of the chapter depending on the nature of the text or due to changes in readers
   2. resulting files must be < 1MB in size, which equates to about 1 min play time.  Therefore, segments must be < 50 seconds.
   3. to be most useful, segments should be about 20 seconds
   4. beginning of chapter must be tagged so that a file will be created for it
3. use File > Export Labels to save the tags for potential later retrieval
4. use File > Export Multiple along with careful naming schema to split the file out into numerous files according to the targeted hierarchical naming schema
5. some renaming of files using the rename utility is still required
6. a final tidying filtering pass is used.  Resulting files are ready for upload to MegaVoice.

Details to follow...

Sample of WAV file naming structure:

Note: every level of file must be present.  If there is not a chapter title, for instance, a 0.1 second silence file must be inserted with that name.

 

5. WAV files sent to MegaVoice

MegaVoice should be contacted in advance of the above processing so that the project can be put in their queue.  While their actual processing turnaround time can be quite rapid, they have a significant queue of projects to be processed, and the lead time required may be several weeks or a more.

 

6. Compiled files ready to load on Ambassadors

Copyright (c) 2006-2008 Joe A. Friberg