I've always considered myself a low-to-mid quality speaker. Like many people, I just naturally learned to talk, forming patterns and habits that have hardened over the years. My limited vocabulary is my most obvious weakness. I rarely expand the words and phrases in my verbal repertoire. But I also am surprisingly poor at enunciating certain words. And I won't even go into my constant use of "umm" filler words...

A few months ago, I started to experiment with a tool that takes a dictionary word as part of a Stable Diffusion text prompt and then outputs a picture. The objective is to speak about the generated picture using the dictionary word. The speech is recorded and then can be judged on various dimensions (tone/words/style/etc). Currently, the tool defaults to peer-to-peer anonymous judging, but there is also a private mode where all the audio stays local to the device.