In order to record your voice using InvTool, you will need a microphone. We strongly recommend a high-quality, head-mounted microphone with a flexible arm. (More specific recommendations will soon be posted on our website: www.asel.udel.edu/speech/mics/.)
First, set up your microphone according to the instructions that came with it. With most microphones, all you have to do is plug your microphone into the appropriate port or ports. With other microphones, you may also have to install some additional software. See your microphone’s instructions for more details.
Next, find out if the microphone is working at all. This is best done by finding the "Sound Recorder" program in your PC.
This will bring up a program that looks something like this:
Put on your microphone, positioning it close to your mouth. Click the Rewind button (the two left-pointing arrows), click the Record button (the red circle), and speak into the microphone for a few seconds. Click the Stop button (the black rectangle), the Rewind button again, and then Play (the single right-pointing arrow) to listen to what you recorded. If the microphone is working, the green line in the Sound Recorder window should fluctuate in step with your voice, during both the recording and the playback, as in the following picture:
At this point, don’t worry about the quality of your recording—recordings made with "Sound Recorder" usually sound poor even when the microphone is set up and working perfectly. Your InvTool recordings will sound much better than these.
If the microphone does not seem to be working at all, try the following:
Make sure you click the "Rewind" button every time before you click "Record" or "Play".
After you are done with the Sound Recorder, you may close it (don’t save any changes).
Positioning the Microphone:
Before the Recording Volume can be properly calibrated, the microphone needs to be properly positioned relative to your mouth. Proper and consistent microphone positioning is essential for obtaining a high-quality synthetic voice with InvTool.
Vertically, the microphone should be level with your mouth. Horizontally, the microphone should be close to the corner of your mouth, out of the way of the puffs of air that your mouth and nose emit while speaking. Finally, the microphone should be positioned as close to your lips as possible without letting your lips touch the microphone when you are speaking. Use the flexible arm on your microphone to move it into the proper position.
Here are some pictures showing the proper microphone placement:
It is important that the microphone be in the same position each time you do recording with InvTool. Perhaps the best way to ensure this is to be very careful when you put on and take off the microphone, taking care not to bend the flexible arm from its original positioning. Store your microphone in a safe place, where it will not be disturbed or played with by other people. If the microphone is accidentally bent or moved, do your best to try to recreate the original positioning.
Calibrating the Recording Volume:
After your microphone has been properly positioned, you need to adjust its recording volume using the "Volume Control" program in your PC:
This should bring up a program that should look something like this:
In the menu bar, select "Options" and then "Properties", as shown below:
In the "Properties" dialog box, click on "Recording", and then "OK":
This will bring up the Recording Volume Control window, which will look something like this:
First, note that this window is organized into vertical panels, one for each type of wave input source. In the picture above, for instance, there are two panels, one for "CD/Line Input" and one for "Mic Input". Your computer may be set up for other kinds of input. In any case, make sure that only the "Mic Input" panel is selected. Unselect the other input sources by clicking on the box labelled "Select" at the bottom of each panel. (On some computers, there may be "Mute" boxes instead of "Select" boxes. Obviously, you want to mute all the wave input sources except "Mic Input".)
Next, adjust the "Balance" meter, which should be set to the center notch (see the picture above).
Next, start up the InvTool program by clicking the InvTool icon on your desktop. This will bring up the InvTool interface:
Click on "Tools" in the menu bar at the top of the InvTool window, and then click "Mic Calibration":
This will put InvTool into "Microphone Calibration Mode". You can tell when InvTool is in this mode by clicking on "Tools" again. There will be a check mark next to "Mic Calibration" in the menu, as shown below:
Now that you are in calibration mode, put on your microphone, and click on the black "Record" button. This will change it into a red "Stop" button:
If you click the "Stop" button, it will change back to "Record".
While the button is red, talk into your microphone at a normal speaking volume. As you talk, notice that the central "Loudness" meter’s needle wavers with the volume of your voice. The white box below the meter gives the volume of your voice, measured in decibels (dB).
Position the InvTool window and Volume Control window on your screen so that you can see and use them both at the same time. Now go back to the InvTool "Loudness" meter. Begin talking into the microphone in a normal speaking voice, at the volume at which you want to do all of your recording. Watch the meter. If your voice gets too loud, the needle will enter the right-hand red region of the scale, and the word "Clipping!" will appear in red below the meter, as in this picture:
This means that the program is removing (i.e. "clipping") the loudest parts of your speech from the recording--something you want to avoid. As you talk at your normal speaking volume, use the click-and-drag adjustment bar in the Volume Control window to adjust the recording volume so that the needle stays within the green and yellow regions of the meter while you speak. The goal is to adjust the recording volume so that it is as loud as possible without entering the red region of the meter.
Test the meter level by reciting the following words: hot, heed, tip, pit. The "o" sound in "hot" is typically the loudest vowel in a person’s speech, and word-initial "t" and "p" are typically the loudest consonants. Make sure the Volume Meter doesn’t reach the red region when you say any of these words. (If you have difficulty with "t" and "p", you may need to position your microphone further from the center of your mouth—that is, closer to the corner of your mouth. See Positioning the Microphone above.) Then test the meter level some more by reciting poems such as "Mary Had a Little Lamb." Continue adjusting the Recording Volume until you can keep the needle in the green and yellow regions while speaking at your normal volume.
The Silence Threshold:
Once you are satisfied that you have set the Recording Volume to the right level, take note of what the "Loudness" meter’s measurement reads when you are completely silent. The meter’s needle should stay all the way to the left, and the measurement should be somewhere between –30 and –70 dB. If it is higher (that is, less negative) than –25 dB, you should take steps to reduce the ambient noise in your recording environment.
Now, take the meter’s average silence measurement, subtract 5 from it, and write this new number down. For instance, if the average measurement was –30 dB, write down –35 dB. Next, click on "Settings" in the menu bar, followed by "Calibration…":
This will bring up a dialog box which looks like this:
Replace the number in the "Auto Trim Threshold (dB)" box (which should be –45) with the number you wrote down, and click "OK".
InvTool automatically "trims", i.e. removes, extra silence from the
beginning and end of your recordings. This setting determines what
"silence": any initial or final part of the recording that is quieter
than the Auto
Trim Threshold is considered "silent", and will be removed. Thus, if
you find that several of your recordings are being chopped off, that
is, if the beginning or end of an utterance sounds like it is cut off
(e.g., begins or ends too abruptly) when you
listen to it, you
should lower the Auto Trim Threshold and re-record.
The Calibration dialog also shows a value called DC Bias. This is a
measure that is constantly maintained and updated automatically as you
use InvTool. Under ideal conditions, silence should be represented by a
signal level of zero. Of course, because there is always some noise in
the background, even when you are silent, the instantaneous signal
level will generally vary slightly around zero, sometimes slightly
above zero and sometimes slightly below zero. Nonetheless, the average signal level should approach
zero during silence. If it does not, your sound card is said to have a
DC bias (DC here stands for Direct Current and means that the sound
card is adding or subtracting a constant voltage from the recorded
speech signal). DC bias values that are close to zero (either negative
positive) are good. A small DC bias is not a serious problem because
InvTool (as of Version 3.2.13) can automatically remove the bias from
the recording. However, if the DC bias value is substantially greater
or less than zero (say, more than 2000, or less than -2000) this can be
a problem. Should you see that the DC bias value is outside the range
from -2000 to +2000, we recommend trying a different combination of
mic, sound card, and (if you are using one) preamp or mixer.
When you are finished calibrating, do not forget to get out of Microphone Calibration Mode by going into the "Tools" menu and clicking "Mic Calibration".
Once you begin to do your recording, you may find that Loudness Meter indicates that some recordings are still too loud or too soft, even when you record them at a normal speaking volume. If you have persistent trouble with this, you may make adjustments to the Recording Volume. In general, though, the recording volume should stay at the same setting throughout the recording process, so try to find an acceptable Recording Volume as quickly as possible. Once you have found this level, stick to it during the rest of your recording. Check the Recording Volume settings before every recording session to make sure they are the same each time.
Remember also to monitor your recordings occationally as you
proceed to be sure the auto trim feature is correctly calibrated. You
can do this by selecting an utterance you have already recorded and
pressing the Play button. You should hear all of the phrase and it
should not contain a large amount of silence either before or after the
speech stops. We'll have more to say about this a little later in the
tutorial (see Section 5).