How can an instrument like the guitar or the piano be made to sound as though it's speaking real words, without any use of a human voice?
The idea behind synthesizing speech from a mechanical instrument is that since speech sounds are (highly complex) combinations of sine waves, and each sine wave represents a pitch, combining instrumental pitches can mimic those same sounds.
This was demonstrated with a piano by Peter Ablinger in 2009 and can be seen in the below video.
There is a very nice video by Andreas Raduege explaining in more detail how this process works. It relates human speech to computer generated speech to piano-generated speech by using Fourier transforms to analyze sound patterns.
In theory, this technique could be applied to any instrument, except for two factors:
- comparatively limited pitch range on many instruments;
- limited- or inability to produce sufficient multiple, simultaneous pitches (this could be overcome, perhaps, by multitracking).
There's a distinction between vocal-like and actual words, and it's tough to get to the latter state.
I played a recording of Clara Rockmore, the first virtuoso of the theremin, and my wife asked me "the woman singing, why is she so sad?" Similarly, in the intro to David Lee Roth's "Yankee Rose", Steve Vai uses a guitar and wah pedal to "converse" with Dave. But there wasn't the articulation to sound like actual words.
There is a way to get closer. There's a device called the Talk Box, which is a small amplifier and a tube. The tube goes in your mouth, and you articulate the sounds. The classic uses I can think of are from steel guitarist Pete Drake's "Forever", Peter Frampton's "Do You Feel Like I Do?" from Frampton Comes Alive, and Roger Troutman on many tracks, including Tupac's "California Love". Listen to them and see if that's close enough.
How to emit actual lyrics without any human voice, only using an object instrument ?
A tool that can create speech sounds without any human voice input is called a speech synthesizer. An example of a speech synthesizer oriented towards musical performance is Yamaha's Vocaloid software. An online search will find many examples of vocal 'performances' rendered with this software (you'll quickly meet Hatsune Miku...)
Like make the guitar or the piano speaking real words, without any use of a human voice ?
A piano and guitar aren't optimal tools to generate realistic human voice - what we can do is to use the piano or guitar as part of a process that creates a sound that has some of the characteristics of the human voice. Typically, some kind of filtering is used to shape the sound's frequency response into formants that simulate the resonances of the human vocal tract.
It is possible to use a formant filter without involving any actual human performance - e.g. it's possible to buy a pedal dedicated to the purpose: