Text to Speech (TTS) with Ardiuno? (Preferably not a shield)

looking to add TTS to a project...
There's Emic2 and dfrobot boards (I'd rather a module instead of a shield, but flexible)
I also spotted kits for SpeakerJet, MagicMouth (but seems dead now), and GinSing. And others based on ISD1790PY

Seems the speakerjet and ginsing are pretty '8-bit computer' variants...
Has anyone put together anything from a chip (or shield/module) and has some sample wav files?

I will be feeding a circuit directly, in place of a microphone, so I'd rather avoid using a speaker directly.

Some chance you may have ideas & thoughts?

I have GinSing shield and the Emic2 module. Emic2 has higher quality speech then GinSing for voice. But GinSing can generate incredible sound effects. The Emic2 module is small, only a 3x4 CM footprint.
As for no shield I have used the TTS library by jscrane. But the quality of speech is low. You can better make out the words if you know what will be spoken.