To add-on to my original post a bit: I have a lot of speech/formant modules. Notably, the Synthesis Technology Circuit Bent E950 and limaflo motomouth formant-filter (driven by a Tiptop Z3000 oscillator); both are outstanding with a rich, high-fidelity audio-output. The others tend to have a screechy quality (e.g., Flame Talking Synth, EMY, etc.), but are also good for certain things. Anyway, taking a speech fragment and slowing it down in Phonogene makes for excellent drone-beds. I'll also use my ISD Sampler for lo-fi slowed-down playback.
Probably my best "speech synth" is the built-in text-to-speech synthesis-tool in macOS using the voice of Siri. Press OPTION + ESC and she will speak any highlighted text. I use this a lot as an input to my ISD Sampler.