Vocals: the final frontier. The human voice might not be as complex as a grand piano when it comes to the number of moving parts, but it is an instrument that’s been around for many thousands of years of our evolution, and we are excellent at noticing anything wrong with another person’s voice.
We’re also much better at noticing a lack of realism in a synthesized voice compared to, say, a synthesized flute. In addition, Vocals are not normally pitched or timed quite in the ways that are convenient for MIDI and our DAWs, especially when it comes to timing lyrics, which means that there are some workflow challenges to solve as well. These things always make a new plugin related to vocals, whether an effect or a synthesizer, interesting.
PPG Phonem (119 Euro) is a new VST/AU plugin that supports both 32 and 64 bit host applications, and its vocals are fully synthesized – there are no samples here. This allows Phonem to provide an enormous amount of control over every detail of the sound and the flexibility to whisper, speak, sing, growl, roar, and even make sounds which are not human at all, but very reminiscent of classic synthesizers.
How Does It Work?
Phonem, as the name implies, organizes sound at the level of the individual phoneme (vowel, consonant or diphthong – a combination of two vowels). A string of phonemes (including any possible pauses for breaths) makes up an utterance, and every phoneme in the utterance has a lot of parameters that can be controlled, from the excitation of the sound generator to how the phoneme blends into the next one. The 12-formant resonator filter also offers control of the frequency, gain and Q values of each formant. The formants are basically what makes a certain noise sound like a human voice – the lower formants allow us to distinguish one phoneme from another, and the higher ones differ between different people. This is all controllable in Phonem, and you can move the parameters around to create any phoneme from any language that you might need, if you know what type of phoneme it is and what its formant frequencies should be.
See also: Alter/Ego FREE Vocal Synthesizer By PlogueThe included phoneme library contains everything that’s needed for American English, German and French (or English with an Inspector Clouseau accent, in case you need that). There’s also a simple dictionary which can convert text to phonemes and make assembling lyrics a quicker process. Selecting the phonemes from the library doesn’t require learning any specific phonetic notation, as each phoneme’s description provides an example of a word it’s found in. Each utterance is saved in its own file, and while a really long utterance could be used for long lyrics, utterances can also be chained together in song mode. There are many different ways to control the pitch and to synchronize the utterances with the rest of your DAW. It does get pretty complicated, but the advantage is that some control modes are better for, say, vowel pads and others work better for sung phrases which need to be synchronized with the rhythm. Preset management is also not simple, with different types of resources managed separately – this allows using the same settings for different utterances. If Phonem was a regular synthesizer, that kind of complexity would be regarded as a drawback, but with plugins related to vocals a somewhat non-standard workflow is par for the course. And yes, it basically means that you are going to have to read the manual.
Phonem features six voices, two female and four male, plus a folder of synthetic textures. Changing from one voice to another doesn’t really make a huge difference – the various other parameters controlling what the voice does have a much stronger effect on what the end result sounds like – but they do sound different and have different strengths. All voices have all the available phonemes, but sometimes there are multiple variations, and a particular variation will work better than the others in a particular word. While testing this, I also discovered that it’s possible to make an utterance which uses some phonemes from the first female voice and some from the second. This sounds surprisingly coherent, actually. In addition to the voices, Phonem can also import wavetables created by other PPG plugins. What this means is that you can create a wavetable of, for example, a violin and run it through Phonem’s formant resonators.
Above the phoneme level, there’s also a big parameter matrix controlling more global parameters such as roughness, aspiration, formant offset, etc. These global parameters can be assigned to two XY pads, four envelopes and four LFOs, along with dedicated vibrato, flutter and growl LFOs. This puts a serious amount of power at your fingertips, even by the standards of non-vocal synthesizers. While other vocal synthesizers make it seem like you’re telling a robot what to sing and how, Phonem feels more like you’re inside the robot controlling every detail of what its robotic vocal tract does. It’s very scientific, in a mad scientist way.
What Does It Sound Like?
With all of these parameters and controls, Phonem is capable of producing all sorts of human-like noises, from whispering to roaring. I haven’t gotten it to scream convincingly, but that might also just be a matter of setting the parameters right. On the other hand, Phonem can also be used to synthesize some completely inhuman sounds, such as bass plucks, for example. While this makes it hard to sum up the sound with a quick description, there are definitely some sounds that are more convincing than others. Comprehensible, realistic-sounding singing of lyrics is tough to get right, as it takes a great deal of tweaking and the vocals still sound pretty robotic in the end.
Phonem is a very small download and while it contains many variations of phoneme data, it does not seem to contain any data specifying transitions between different phonemes – something that other vocal synthesizers use a lot of. All transitions in Phonem appear to be calculated on the fly, which provides more flexibility, but makes natural pronunciation harder to achieve. Perhaps a future version of the instrument can develop a “best of both worlds” approach. However, for the time being, if you are looking for an alternative to Hatsune Miku which is easier to use, Phonem is not it. Another thing that’s possible in Phonem is classic synth leads, pads and basses. These are quite good, but again there are other synthesizers out there which make generating those sounds easier.
However, getting away from comprehensible lyrics is where Phonem really shines. Choral pads, “yeah” or doo-wop vocals can be amazingly realistic, organic and expressive. Tibetan throat singing can be imitated quite well, too. More abstract vocal textures with nonsense utterances, especially repetitive ones sweeping back and forth across the utterance, are also a definite strength. If you need singing sci-fi aliens, or even growly singing dinosaurs, that’s quite easy to do. You can make a nice polyphonic patch doing something weird, assign some parameters to the XY pads, hold down a chord and just play with the XY pads for several minutes making the sound change and evolve. For something more commercial, short weird utterances make an amazing alternative to currently popular chopped vocals, especially if you turn up the excitation on a phoneme which isn’t played first. This way, if a MIDI note is held long enough to reach that phoneme, the sound suddenly turns metallic when it clips. Also, thanks to that metallic quality and the ability to control the length of phoneme transitions, utterances of vowels slowly morphing from one to another can make great dubstep growls.
A lot of the factory presets make me think of classic Krautrock. On one level that can’t really be true, because Phonem is completely different in technology, much clearer in sound and much more powerful than anything those guys had available in the 70s. It does, however, have a similar vibe, and many of the sound designers who created the presets probably like Krautrock. Again, the words “mad scientist” come to mind. Similarly, most of the music Phonem users have made in the first few weeks since its release is on the abstract soundscape side of things – I haven’t seen any pop song covers yet, though I suppose Phonem could do an amazing job of singing Hanson’s “MMMBop”.
The Verdict
In some ways, this is a vocal synthesizer that is most different from Hatsune Miku and Vocaloid. There are no virtual characters with a very distinct voice, image and personality. Also, while Miku is mostly known for singing pop songs, Phonem is best at synthesizing abstract weirdness.
Phonem is very powerful and provides a level of control that no other vocal synthesizer does. It’s not the best vocal synthesizer to use as a substitute for human singers singing actual lyrics. It is, however, great at imitating human singers who are not singing comprehensible lyrics, and even better at many vocal-like things that no human singer would be capable of doing.
More info: PPG Phonem (official product page)
PPG Phonem Review
Phonem is very powerful and provides a level of control that no other vocal synthesizer does. It's not the best vocal synthesizer to use as a substitute for human singers singing actual lyrics. It is, however, great at imitating human singers who are not singing comprehensible lyrics, and even better at many vocal-like things that no human singer would be capable of doing.
-
Features9
-
Workflow8
-
Performance10
-
Design9
-
Sound9
-
Pricing9
