As a virtual vocalist created from the voice data of a female death metal vocalist, Marie Ork is a little unusual. In addition to singing, she can growl and create a very wide variety of strange sounds. The Marie Ork voice bank contains three voices: clean, growl and space. All recordings were made for English, but Japanese lyrics are also supported, and the clean voice is capable of singing in Latin, using the phonetic table found in the Latin_X-Sampa_guide_for_Marie_Ork.pdf file.

The Alter/Ego user guide can be downloaded from http://www.plogue.com/downloads/ the information in it applies to the clean and talker voices, while mostly applying to the growl and space voices as well.

THE CLEAN VOICE

This is a voice usable for "normal" singing, and works essentially like most Alter/Ego vocals. It was recorded fairly loudly, and slowly, so it works best for slow passages. For typical pop songs, raising the wave rate slightly is often a good idea. She sounds most realistic in the standard classical alto range, but can sing much higher or much lower. Karoryfer Samples generally recommends using expression (CC11) to control volume and a small percentage (10-15%) of HF Sizzle, while velocity controls 0% volume and 100% glide time. With longer glide times, at least 60% glide and 70% legato glide, this is a good way to create expressive and organic vocals, especially with live keyboard input while manipulating CC11 for dynamics and CC1 for vibrato. However, that's not the only way to use the vocal - it's merely the developer's personal preference. Some of the presets have "exp" versions which are set up like this, and "vel" versions which sound similar but have velocity assigned to volume.

THE TALKER VOICE

This is the clean voice operating in Alter/Ego's talker mode. It's not a dedicated text-to-speech voice, and more realistic speech can be achieved by using the clean voice with high glide times and individually timing every syllable.

THE GROWL VOICE

This is a death metal voice which produces (almost) no discernible pitch information. Therefore, it can't be used for singing melodies, and pitch vibrato has no real effect on it. There's generally not even any need to keep notes in key. However, that does not mean notes make no difference - lower notes will be somewhat deeper and higher notes more screeching. For metal vocals, it can be useful to have the last few notes of a phrase lower or higher than the rest, to add a bit of variation and expression. Differences in volume can also be useful for adding expression, ideally by varying volume, or linking expression to volume (as recommended for the clean voice) and varying that. Rising screams can be created by a series of legato notes with increasing volume and pitch.

This voice can be forced to produce a discernible pitch with extremely low wave rate settings. The growl throat sing preset is an example of this. Generally the space voice works better for producing pitched notes with harsh textures, however.

THE SPACE VOICE

The space voice was an accidental discovery - Alter/Ego is capable of detecting pitch even in very noisy death metal vocal data and synthesizing a pitched sound based on that. With certain settings, the HF Sizzle control becomes a de facto pitched content control. Low wave rate and low phoneme speed also make the pitch clearer.

A discernible pitch does not make the lyrics any clearer than the growl voice, and still sounds rougher and noisier than a normal singing voice, but it opens up all sorts of weird possibilities. Most of the presets using this voice are in abstract, experimental and just plain weird territory, though it can also be used for semi-melodic folk-metal-ish growls and for morphing from a somewhat noisy vowel to a growl.

PHONETICS

The clean voice includes standard English phonemes plus a few extras. The "a" is an open a, used in Latin and Japanese, and also useful for wordless "ah" vocals in English. The "r" phoneme is a rolling r, used in Latin. It works properly only at wave rate speeds of approximately 80% to 120% - beyond that range it becomes a sound effect rather than a recognizable r.

"4" is a shortened rolling r, used for the Japanese r, as well as in certain English words with an American accent. For example, "4" can be used as the "tt" in "butter" or "t" in "waiting". There is also a "ts", similar to the German z in Zug, which is useful for Latin. Finally, there is a "h*2" phoneme, or alternate h, which is softer than her regular "h". That makes it useful for things such as "aha". This phoneme does not have transitions to and from all other phonemes recorded, which means it's useful basically only between most vowels and the consonant m, making it usable for both "hm" and "mhm".

Only phonemes found in Alter/Ego's standard English dictionary are recorded for the growl and space voice. There are no additional American English phonetics. No Japanese or French phonetics were recorded except for "a", but the nearest English phonemes will be substituted for the X-Sampa phonemes used in standard Japanese and French. So, Japanese and French will have an accent, though it might not be so easy to hear it in the growl.

Alter/Ego is more fastidious about pronouncing words than most vocalists are in real life, and therefore some editing of the phonetic data in X-Sampa mode may be in order to sound more colloquial and less academic. This is obviously especially true for death metal, where skipping some consonants in the middle or even at the ends of phrases can be helpful, as can substituting some vowels with lower formants, for example u for O, or O for @ or V.

THE CHARACTER

She's a cyborg, a goblin and a witch, which explains her cybernetical-magical ability to produce so many different sounds with her voice. Her favorite human vocalist is Edith Piaf. We didn't bother to think about stuff like age, height, weight... We did think about whether she has a bellybutton or not, didn't actually decide, and then several people drew her with a bellybutton, so she does have one after all.