Laying in parts

I was going to write in about the same problem. No matter what I do, it is very difficult to work with the punch in and out.

For punch in and out, It is important to select a span of at least minimal duration. Selecting a position only will replace everything that follows it.

Figure recognition should never replace anything outside the selected span. If it does, it's a bug. The only exception is when segments overlap the span's bounds.

Editing a figure does not change the take it was derived from. The original take still sits in the phrase. So, if you select a span and press "Apply" once more (i.e. run a new figure recognition), the selected portion of the figure will be built from scratch and the edits are lost.

I noticed that sometimes there can be a display refresh missing, i.e. the figure created after a new take does not immediately show up. Clicking on the current container fixes this (a container selection refreshes almost every view entirely).

Andre

Andre , I did select a span and it still erases everything after it.I made sure no notes from the next measure were in the span boundaries. I think I am doing this correctly . When I hold down shift and record I can make it work but I thought that shift record was for merging new notes into the same measure. In one test however, I deleted the notes in that measure then used Shift Record. It recorded the new notes but it also brought back the previous notes I deleted . I hope I am making sense.

@tokyorose: You can edit the take and listen to it using the "Probe" button. That will render the parameter as is (not the phrase). This btw works for a couple other parameters too.

FYI: As figures are derived from takes through figure recognition, Synfire retains all takes in the phrase (hence the parameter "Take"). A punch-in recording first replaces the span in the take, performs figure recognition on the whole take again and finally copies the span of the newly recognized figure to the Figure parameter.

The take parameter collects all your takes over time. Hence previous takes have an influence on subsequent recognitions. Example: If you punch in at the middle of a melodic run, Synfire might extend a previous segment that started outside the span.

Perhaps that's just too ambitious and we should probably make it simpler, i.e. perform recognition inside the span only and ignore everything outside.