Westlake village in the movies. Demolition Man, Encino Man, kind of made 1993 the year of the upscale Calif* suburb. It might have been kicked off by Father of the Bride or it might have been the wave of yuppies that began under Reagan, becoming the upscale Calif* suburbians by 1993. Those same neighborhoods no longer house the middle class but 1% venture capitalists, gootube stars, C suite executives.
https://growmane.blogspot.com/2024/04/taxes-rent-bills-give-april-12000-tag.html
Anyways, it got lions comparing themselves to Ego Monger. She's still in a much less walkable area, in lion opinion. Westlake village is heavily fenced off from the ruffians next door so all travel is on 1 big ugly blvd. The SUV was king in 1993, but walkability is king, nowadays.
Ego Monger's husband may have achieved the dream: a $1M house near Westlake village, but lions still managed to rent a tiny apartment in a similar upscale Calif* neighborhood for 25 years. That's something.
Mr. Monger's interest was nearly zero because of covid relief funds. Don't think any covid relief funds relieved anyone but gave homeowners double their money.
-----------------------------------------------------------------------------------------------------------------
Of all the pieces of software commonly in use, obviously the web browser is the most used of them all, but the humble video editor is probably #2. The web browser is kind of an operating system, abstracting every function in a common language.
Lions can't think of any other piece of software dedicated to a single purpose, that's more constantly used by every living being on the entire world, more beaten up, than the humble video editor. It's quite an evolution from 25 years ago when video editors were virtually unknown in daily life. The amount of abuse they take is kind of a vindication of the workflow they've all standardized on.
The way effect configurations are represented in 1 big pane with all the effects stacked in a column, the organization of media into clips, the way audio & video are locked together, the standard drag & drop movements, the editing of curves with easing presets, are all as proven as breathing.
Lions, unfortunately, have never been comfortable with drag & drop editing or clip objects with locked video & audio. Locking would have to be set for each edit, with an option to automatically lock everything in a highlighted region. It needs a way to indicate which edits are locked & with who.
The user would have to select based on edit. Well, CTRL already enables free dragging but it's sticky. CTRL without dragging could enable selection of an edit. So you would select a bunch of video & audio tracks, then call a lock edits function. The relative startprojects of those edits would now always be the same.
Currently, it just grabs all the edits with a matching startproject. Locking would grab all the edits with the matching locked property.
It would draw some kind of indication, maybe a different colored outline. Unlocked edits would have their own outline. It's probably time to start drawing minimal 1 pixel edit outlines, no matter what.
---------------------------------------------------------------------------------------------------------
Curve easing is relatively new, but seems to be the way it's going to be. It basically sets 1 or both of the bezier controls to the tangent of the curve so there's no sudden change in the slope. Then the control points automatically adjust as the slope is changed. Something like that could probably be added to the floating point keyframes.
Right now, it has somewhat of an easing type in the mode menu. The problem is we want the slope before this keyframe to be steeper than the slope after the keyframe with no sudden change. It requires the control points to form a tangent & it's not possible to do that.
Locked bezier can get a close but never a perfect tangent & the slope at the keyframe can't automatically change to follow the slope of the line. You have to adjust the control points every time the line moves.
The most automatic values are 0, but that only works if 1 side of the keyframe is slope 0 or the center value is a minimum or maximum. Another mode called ease or tangent might do the job. Quite a few places would have to recompute the control points. It might need to privatize the value & build the control point computation in the value setter. It would need to recompute after cutting & pasting. It's looking like a nest of global recompute functions, but they always have missed corner cases.
Another way is to privatize the control points & recompute in the control point getters. Then, there's just neglecting the control points except in rendering & drawing the curves.
FloatAutos::automation_is_constant, FloatAutos::get_value are the only functions which depend on the tangent calculation. The control point variables are not interactive & left at whatever the user had before tangent mode.
After getting far along with it, found there's really no automatic solution.
Locked bezier achieves a smooth curve with any slope. The most useful algorithm might be an iterative curve fitting which tries to achieve the most gradual transition into the neighbors without overshooting. It would have to test every intervening sample for overshoot, over a logarithmic range of control point values. It would have to be done for every keyframe move, copy, paste, clear.
Grok spits out a ponderous description of some kind of iterative process which generally sounds like the same thing.
The curves plugin has its own calculation to compute control points. It doesn't eliminate overshoot, so it might benefit from the same iterative curve fitter.
Cinelerra multiplies the control points by 2, so the timeline control points for a smooth curve never match the slope of the line.
Unfortunately, the smoothest transition at the neighbor points doesn't always lead to the smoothest line through the center point. It really depends on where the center point is. If the center point isn't exactly halfway between the neighbors, the neighbor points have to bend more if the center point is to bend less. It would have to target maximum overall smoothness & there's really an infinite number of solutions.
Everyone else is probably just locking the control points to 0. There could be an ease mode which does that, but it's overkill since they default to 0.
----------------------------------------------------------------------------------------------------------
Reviewing the recording code for error propagation, it still has bits of loop recording from 1997. What a huge wasted effort. That was only ever used for continually recording the radio on a hard drive with only 4 hours of storage & picking out songs. It would never work for any compressed format, yet somewhat of an implementation of it continued through the addition of batch recording. Young lion thought a video looper would eventually be based on a chunk algorithm, but the labor requirements for every original vision eventually reached infinity. Storage became so limitless, looped recording would be worthless now.
Lions always debated splitting Cinelerra's recording into a separate program. An audio loop recorder would definitely be a separate program & it would use its own file writer instead of a library.
Anyways, young lion for some reason was always setting write_result to 0 inside RecordVideo::write_buffer & RecordAudio::write_buffer. There was an even older bit which overruled the audio failure only if video was being recorded.
Those were all taken out. It's possible older file writers were prone to transient errors from bugs, missing return values, or uninitialized variables. It's possible the recorder was supposed to be error resilient if a single frame overflowed its encoder buffer.
Comments
Post a Comment