I use Audio Bend to quantize guitar bass and vocal tracks routinely.
The audio bend detection algorithm should have separate modes for picked instruments and vocals.
When quantizing, the portion of the audio that needs to align with the grid is the vowel - or the note, not the consonant letter or the pick attack. Piano and similar instruments do not have the pick attack but instruments similar to guitar 99% of the time have this and this needs proper editing otherwise it sounds dragging behind.
If the algorithm could detect pick attacks, sh, t, k, ch, s sounds and place the bend marker right at the start of the actual note (vowel or instrument note) that would save a ton of time. Currently I have to manually go through each track, place the marker appropriately and then hit quantize.
Also the auto detection sometimes places markers really close, and when quantized they open an unnatural gap that causes a hot red portion with sounds distorted. There should be a threshold setting that would help prevent markers to be created too close to each other.