[dodo53]

I wonder if there's a way you could help this with software. I remember reading a long time ago about singing pitch training by realtime headphone feedback (like - if you're singing flat, you are feed back your voice distorted even more flat[1]). I wonder if you could detect and exaggerate differences in prononciation.

[1] - closest I could google now was http://www.ncbi.nlm.nih.gov/pubmed/8173617 which makes it look pretty ineffective actually :o/

[mietek]

Wouldn't visual feedback be easier to comprehend?