RealSinger

To learn how Real Singer works, let's quickly create a new Real Singer voice, for Latin songs. Why Latin? This language has few distinct sounds, so you'll be able to record quickly. Also, there are a number of public-domain Latin hymns that you can use to test the voice.

What you need

You must be able to record your own voice to your computer's hard drive. The simplest (not necessarily best) method is to use a microphone of the kind designed to connect directly to your computer's sound card input. However, microphones for professional audio usually must be pre-amplified before your sound card can record the signal. Or, you can use an external recording device, and send its line-out or headphone signal (never the loudspeaker signal!) to your computer's sound card line-in. You can even do the recording with another computer, then transfer the digital audio files. But for now, we will assume that you are recording your live voice directly to your computer.

Let's record!

When you are ready to record, find a quiet area at a quiet time. - On Melody/Harmony: create a new, very simple document with just one staff. To the left of the staff is a set of icons, shown here. Click the black triangle to open the staff-specific menu, and choose "staff with lyrics." If the Virtual Singer stage does not open automatically, - On PDFtoMusic / PDFtoMusic Pro: load a PDF file that includes lyrics in latin, e.g. "Ave Maria" in the "PDFtoMusicDemos" folder Then select the Virtual Singer palette in the "Windows" menu.

The Virtual Singer palette is a stage, shown here in reduced size. Standing on the stage is the picture of your singer, and underneath his picture is a question mark. Click the question mark, and choose "Latin" from the available languages. The question mark changes to an icon. Then double-click the singer. In a few moments, a dialog box appears. Click "Real Singer".

A new dialog box appears. Since you will be recording live, click "record." Be quiet while Real Singer measures the background noise. If it is too loud, try again. After Real Singer measures the backgrund noise, it asks you to say "aah" for several seconds. It is measuring your natural speaking pitch, so that you can listen to sample words at that pitch.

If Real Singer insists that there is too much noise, but you don't know why, possible solutions are discussed in advanced recording techniques. For now, you can bypass thee noise measurement by choosing "set" instead of "record." If you do that, then the next dialog box allows you to manually choose the frequency at which words will be presented. A range of 90-130Hz is comfortable for most men, double that for most women.

The Real Singer recording palette will appear. At the left are a list of words that you must record. Real Singer will offer you the words in order, unless you un-check the "automatic" box. You can save an audio file of each recorded word by checking that option. At top is the current word. If you wish to hear it without recording, click "listen to the word." When you wish to record the word, click "get." Real Singer will play the word, then you will repeat it. Try to imitate what you hear. If your voice is too loud or too soft, get the word again.

If your voice is OK, Real Singer will process the sound and identify the phoneme. The first Latin word is "quid," and Real Singer is looking for a drawn-out "u" sound. At the right is a reduced-scale image showing what Real Singer found for a sample recording. The bright area in the middle is the identified sound. The dark curtains at left and right block off the "q-" and "-id" portions.

Sometimes, Real Singer cannot find the phoneme, or locates it incorrectly. You can move the "curtains" by dragging them with the mouse cursor. You can play back the whole recorded word, or just the selected portion. It is important to choose phonemes carefully - these are the building blocks used to construct all other words! More information about how to adjust the selection range is found in the "Adjusting phoneme selection range" chapter. When you are satisfied, click "validate" to add the processed phoneme to the Real Singer voice.

After you validate a word, its appearance on the word list changes. You know which words you have validated, and which ones remain. If you are not using automatic recording, you can choose words in any order, or re-record words that have already been validated. If necessary, you can save incomplete results, and finish the list at some other time. You can even leave some words un-recorded, and Real Singer will use synthesized sound from the Virtual Singer database as substitutes (not recommended). For best sound quality, you should record all of the words, in a single session, so that your own voice is consistent from one word to another. After you have finished recording your voice, close Real Singer to return to the Virtual Singer dialog box.

By default, Virtual Singer asumes that your voice is male, and that you will use it for notes on the Treble staff. Therefore the sound will play one octave lower than the notes are written. If you are female, or if you are going to use this voice on the Bass staff or with Treble-8vb, then set the octave shift to zero. You can also change the singer's stage appearance, if you wish.

While the Virtual Singer dialog box is open, click "Add To List" Give your voice a name and save it. Then move it in the Real Singer Latin voices folder that you created by right-click. This name identifies the voice file, not the character shown on stage. The saved voice file can be used by any singer of the same language, in any other documents. The stage character always has the name of the staff he sings, rather than the name of the voice file he uses.

Click OK to close the Virtual Singer dialog box. You are now back in your music document, with the stage still showing. Save your document, even though it does not have any music.

Using your new Real Singer voice

Open the sample Latin song "Exsultate" in "VirtualSinger/Demos/Latin" subfolder. There are two singers, "Ron Real" and "Vic Virtual." Double-click the image of "Ron" to open Virtual Singer. Ron starts as the default male Virtual Singer, but you will change his voice. In the menu of voices, find the Real Singer voice file that you just created (remember, you put it in the Latin folder). Virtual Singer may tell you that this change will lose the previous setup. Confirm, then click OK to return to the stage.

"Vic" sings with the default male Virtual Singer voice. Leave "Vic" that way, unless you are female and want "Vic" to sing with a female voice. If so, double-click "Vic," and select "Soprano." This voice is not in the Latin folder, but Virtual Singer is multi-lingual. Confirm the change of setup, and click OK to return to the Virtual Singer stage and music.

Save the file (save as) with a different name. Now, play the music. Real Singer must pre-calculate the voices, so there will be a delay. "Ron" and "Vic" will sing together in harmony, with "Ron" the higher voice. You can mute one voice or the other, if you wish.

Most likely, your "Ron" voice is rough. Double-click "Ron" to open Virtual Singer, then click "Edit voice." A new dialog box appears. Under "timbre," choose a value near 30 for each of glottal and opening. Re-play it, and hear the difference in "Ron's" voice.

If you save the music, it will retain the changes made to "Ron" for this music only, without changing the voice file you recorded. If you want the changed voice to be available to other music, within Virtual Singer you can "save preset" to a new voice file. Don't try to use "Ron" to sing English, or any language other than Latin. He doesn't understand it, because you did not record most of the phonemes needed to make words in another language.

When you play your Real Singer voice, you may notice that some phonemes are too loud, or too soft, relative to the others. If a phoneme is not consistent with others, you will hear the same problem each time the phoneme is used. To fix this problem, open the Virtual Singer palette, double-click on the singer, and choose "edit voice," then the "advanced" tab. Choose "edit phonemes." Find the offending phoneme in the list at the left, and select it. On the right you will see several vertical sliding controls. The rightmost two control the phoneme volume at start and peak. Move these up or down, as needed. Play your music again. When you are satisfied with the relative volume of the phonemes, save the voice preset. There are numerous other phoneme adjustments available.

Congratulations on creating your first Real Singer voice! The rest of this Real Singer documentation describes ways to improve your voice recording, ways to improve the precision of the voice fragments, and ways to use Real Singer adjustments to improve the recorded voice.