Monday, January 5, 2009

logical / heirarchical editing in FCP

I'm editing a music video. I have dozens of camera angles from a dozen performances of the song, shot at various locations.

Each angle is a file named by event and source "warper - elsa" was shot by elsa at a performance at the warper event.

I make a "clip sequence" called "warper elsa" which gets that single angle, time-cropped to the relevant song. I double click the instance of "warper - elsa" file in the "warper elsa" sequence to open it. I then fix up anything on the clip level - color correction, scale and position for fixed cameras that are too wide, perhaps automate zoom and light balance to counteract any shortcomings in the raw footage.

Each performance of the song is slightly different, this part or that extended for extra bars. Therefore, I create an "alignment sequence" named "warper elsa aligned" with the studio track. I drag the clip sequence "warper elsa" into the alignment sequence "warper ela aligned" to create a reference. I then cut up the clip and align it with the studio audio track, matching verse to verse and breakdown to breakdown.

After repeating for each angle, i have a bunch of alignment sequences which are all synched with studio track.

Now I create an "all-up" sequence. I embed all the alignment sequences into the all-up sequence. I ctrl-click each instance of an alignment sequence and "open in viewer", then use the motion tab to move it into a 9-up position: x = +-245, y=+-160. Too many angles to play more then a second in real time, but i can render it and it serves as a nice reference, a way to view all angles when looking for edits.

I also create one or more "edit sequences". Edit sequence also gets an embed of each alignment sequence. But rather than using motion for all-up, I cut out bits of the sequences to reveal each in the appropriate places.

Now I have an end-to-end project for editing this song.

The idea is that if two angles have excessively different color balance, I can color correct their clip sequences. If the mouth doesn't quite match the voice, or the headbanging starts two bars before the music cue, I can open the alignment sequence for that angle and nudge the segments by a few frames or swap bars.

PROBLEM: if i double-click a clip sequence in the browser to open it in the timeline, then open the underlying clip in viewer and desaturate it, it appears b/w in the clip sequence. However the alignment sequence stays in color. Double-clicking the instance of clip sequence in the alignment sequence timeline, it opens in the same tab as clip sequence - except without the desaturate filter. Here it gets a little wierd - the instance of clip sequence embedded in alignment sequence is not a reference, it appears to be a copy. It may share a timeline tab with the original clip sequence, but it is a copy - made at time of embedding presumably.

IT GETS ODDER: So if dragging one sequence into another makes an embedded copy (rather than reference) then any changes made to the alignment sequence should not be reflected in the all-up sequence. But in fact, testing with a desaturate on a different angle, all-up *does* honor filters added to instance of clip sequence in alignment sequence (although i have to re-render before I see it).

The only obvious difference is the level of indirection - in failure case desaturate is applied to an instance of a clip in a sequence - in success case the filter is applied to an instance of sequence in another sequence. If I put each clip in a "file sequence" then embed that in the clip sequence, would that be a workaround?

Test: the alignment sequence has a bunch of cut up bits of the clip sequence - except a copy of the sequence. Is it one copy or many? Double-click a chunk of clip sequence in the alignment sequence to open it in a timeline tab. Add a desaturate to the instance of clip in this (copy occupying the same tab as original) clip sequence. Back to the align clip, it is honored. Honored also in the all-up.



Current strategy: delete the clip sequence, solve clip-level problems in the tab that open when double-clicking instance of clip sequence (which isn't lost even after clip sequence is deleted from browser) in align sequence. This reduces the clutter in the bin, a nice side effect.

I have trouble rerendering all-up seq after filtering clip seq. Sometimes rendering works. Fix: toggle visibility,

Double check: filter in original clip seq, force update in align. Nope, still not honored.

Note: fixing align seq, def honored in both edits. so, that is workable.

According to this page, when 'nesting' changes made to the original sequence are reflected in the sequence which nests it. http://www.scottsimmons.tv/articles/nesting.html


My all-up and edit sequence strategy is working well. But to edit efficiently i need to have the edit seq open in timeline/canvas, and i need to be able to view the all-up seq in sync. Multiclip editing allows this, you can set the viewer to 4- or 9- up. But multiclips are limiting (you can only use clips, not sequences, so there is no nesting of source angles). But i can't find a way to play a 2nd sequence in sync:

can't make 2nd viewer or canvas windows

if playhead is open, then viewer automatically loads clip/seq currently visible in timeline. I need to view the all-up constantly, in sync with my current edit. I wish i could make playhead sync follow the time without changing the clip... that feature is foiling my plans!

Tried to put the all-up on top in sequence (so it would auto-select in viewer with open sync) but find a way to keep it from being visible in canvas. Can't use filter because those show up in viewer. Try changing composite mode, but that is sloooow.. plus all modes have the top layer visible, plus the viewer shows the same composite as canvas, not just the top layer!

Best workaround: put the all-up seq as top layer in edit seq, make that top layer invisible, open in view, viewer playhead gang sync. You can't watch both at same time, but you can play all-up (in viewer) or edit (in canvas) back-to-back without any other gui clicking.

NOTE: edit seq has only one copy of audio because two copies is too loud. Result is that when opening instance-of-all-up-in-edit-seq in viewer, it plays silent. Bit it is possible to drag the all-up seq from browser to viewer, as long as sync is gang it'll catch up to activity in the edit timeline and be playable on its own with audio. You have to manipulate timeline and click play in the appropriate window, and you can't watch both angles in sync, but this does allow you to view the preview track and exit back-to-back. You can use ` to create matching markers on the viewer and canvas so you can play back the track from the same moment.

NOTE: there seems to be some funny business, relative-position-matching. To get things started in sync i had to put both viewer and timeline at (synchronized) zero, then set gang, then play in the timeline.

WAIT: in gang mode viewer will catch up with the canvas, but the canvas won't keep up with viewer - viewer plays on its own. Doesn't lose sync in the sense that playing the canvas at different location will cause viewer to catch up. But even with viewer and canvas both ganged, canvas will not follow viewer.

NOTE: "Liberty Edit" seq had one instance of ref to "flow ilan line feed" but that was cut up a bit for creative purposes. It seems like 90% of the cut up bits open one copy of the seq, one bit opens a fresh copy. Even after fixing black levels for all the hand segements of line feed, one bit of edit showed black problems. Double-clicking this segement in edit seq opened "flow ilan line feed" copy that didn't show any of the edits i had previously done to set black levels. The problem segment wasn't at -100% speed or any other conspicuous attribute.

No comments: