Ptichi is being built for a very ordinary problem: you know what you mean, but the other person has to work too hard to reconstruct it.
The method is deliberately small:
Take A → listen → change one thing → Take B → compare → try it on new words.
That is less glamorous than asking an algorithm for your “confidence score.” It is also much closer to practice.
Why two takes
A single recording gives you a sample. Two deliberately different takes give you a comparison.
The useful question is not “Was my voice good?”
It is:
Did the change make the intended point easier to hear — and can I reproduce it without copying the exact sentence?
That last step matters. A polished second attempt is not the same thing as learning.
What Ptichi is not trying to do
Ptichi is not designed around a universal voice-quality number.
Acoustic observations can be useful. They can show timing, energy or other measurable differences when the capture and analysis support them. They do not automatically reveal confidence, competence, personality or whether someone is a “good speaker.”
A precise number can still be a precise answer to the wrong question.
Current status
Ptichi is under active development.
The public site will document product capabilities, limitations and release state as they become real. It will not invent a download button for a build that is not ready.