Which phoneme-to-viseme maps best improve visual-only computer lip-reading?

Re-assign community