Many contemporary sci-fi projects give pretty natural speech to robots because technology is assumed to have achieved it. One particular recent BBC project was interesting because it shows a slightly flawed implementation where totally natural speech is occasionally interrupted by a mechanical stutter – just a fraction of a sec and rarely but enough to remind you and the characters that this is not a normal human.
It would help to know what kind of robot you want. Does the robot have any personality? Is it supposed to be sophisticated, to emulate humanity or not so much? Is it threatening, comical, somewhat broken down?
If broken down I would use a glitch plugin to loop some of the words, varying the buffer size for all kinds of stutter and grinding effects on whatever process you’re using.
If the robot can be comical, small or obviously fake human, you could use mda vocinput, a free plug-in and apply a random automation envelope to the pitch parameter or use the track function so that it will track your voice. I have had some good results with this.
Vocinput provides a sound that you can use with MDA’s vocoder or any other, so you could apply whatever effects to that sound you wanted in addition to the tracking/randomized pitch.
Zynaptiq Morph would be excellent for this as the possibilities are endless. Their Adaptiverb also provides some possibilities because of the cross-filtering capabilities. You can make a square wave talk etc.
there are probably other tts engines out there you could use and modulate. Hope this helps.