It would be really useful to have an option for non-uniform quantization strength, best of all via linear (log?) plot, so the performed shift would depend on the deviation of current item/note from grid.
Two use-cases (resulting shifts pictured are not exact!):
1.
Straight dependence. The further the object is from grid, the more it's shifted. (q1 attachment example)
This results in gentle correction of tempo drift which saves the distinctive character of performance but tightens up the time.
2.
Inverse relation. The further the object is, the less it's shifted. (q2 attachment example)
This may be useful when there's an assumption that the performer intentionally "bent" the time (e.g. played behind the beat) in a musical way
some of the time, or used some occasional subdivisions not in the grid (like playing a bar of quintuplets).
PS: Would be great to see this in a
QuantizeTool by @mpl as well.