"Dear Nonlinear Picture Editing Department"

by Larry Blake

Dear Picture Department:

I suppose that those of us in post-production sound shouldn't begrudge the world of picture editing the sense of freedom brought about by nonlinear, hard disk-based editing. Being freed from the shackles of splicing tape and inflexible tape-to-tape offline editing has to feel at least as liberating as it did to us. (Although I hear you have a harder time than we do in justifying to producers what, if any, extra costs are involved. I'll try to touch upon those aspects when possible.)

What follows are a series of suggestions to help your picture and my sound departments interface as smoothly as possible, keeping technogrief to a minimum:

•  Establish a dialog with your production mixer and with post-sound in advance of shooting. I know that producer types sometimes don't want everyone talking; it's amazing that they have no idea of how much money we could save them if they would just let us talk among ourselves. What could they possibly view as the downside of such a dialog?

It is crucial that we all agree and understand the path that picture and sound will take on their way from shooting to picture editing to sound editing to mix to answer print. And when it's agreed on a path to take ...

•  Do a sync test. This is real simple. Shoot a continuous 9-minute take with someone talking, holding a timecode slate and clapping the sticks about once every two minutes. Then do two other short takes with the remaining film left in the camera.

Have the picture and track printed twice each: one set will remain uncut as a bible roll for your telecine house; the second set will be conformed after you do a test edit, intercutting among all three takes. (This is a simplification of the process, but I think you get the point.)

The best time to shoot a sync test is in the few days right before shooting begins, when the camera and sound teams are checking out equipment and the editorial department has just been hired. This will allow you to limit your variables because the chain will involve the production camera, the production recorder, the telecine facility and the nonlinear editing system.

While such a sync test will solve a multitude of post-production problems by telling us that your numbers and pictures do (or do not) match up, it will be of even greater help to you in providing a sense of security that the shooting  / telecine / editing / conforming process really does work. Just like the old adage "a watched kettle never boils," a non-tested nonlinear system is never up and running in the first week of shooting. While so many problems end up being real stupid and simple (software version problems, they didn't ship the right cables, etc.), just as often they are fundamental and serious (like the telecine facility doesn't have a clue as to what they're doing). And, as always, please observe Blake's Third Law of Disagreement: Don't let anyone tell you that no one has ever complained before or that this is what they do for James Cameron. Repeat after me and tell them to read your lips: "I am complaining and my name is not James Cameron."

In the days of 35mm sprocketed picture and worktrack, sync was taken for granted, with clunky old clapsticks providing an absolute sync reference that was preserved by inked edge-coding. In those not-so-olden times, when dialog editing would begin, we would receive the spliced work track, indicating exactly where edits were made in addition to the source scene and take of each piece of the track. Also, any 1:1 copies of the work track were in exact sync with the original; there was never any doubt. Today, there are many more paths in the road to cutting dialog and many more places to screw up. Sometimes there's still mag used solely for the purposes of interlock dailies, whereas at other times there's no mag or workprint at all, with the picture coming from a negative-to-tape transfer.

The constant variable (?) here is the telecine process, one which contains many minefields for the unsuspecting, even when it involves already-synched picture and worktrack. "How can this be?" you might ask. Each telecine bay involves varying amounts of video delay due to processing (such as noise reduction) between the scanning device and the videotape recorder. A good facility will have quantified said delay and will adjust sound accordingly, either by offsetting start marks or with a digital delay line (the latter of which is fine for dailies, but I would be wary of doing any sort of mastering this way). But many facilities don't take this into account, and you can conceivably leave telecine with all of your audio 1.5 frames ahead of your picture — this when what you delivered to telecine was dead-nuts in sync with start-marks on mag.

At least the above example would result in a constant, quantifiable error that could be corrected. A more sticky problem occurs when transferring sound to picture either during or after telecine, most frequently by reading timecode numbers on slates. There are many variables, not the least of which is the 30 fps video/24 fps film speed dilemma and the fact that there are often subjective decisions being made by people reading blurry timecode slates in the middle of the night. (Or, perhaps even worse, by an assistant editor synching dailies to a none-too-sharp digitized image.)

The net result is that a take which is digitized out of sync will stay out of sync forever, unless a sharp-eyed dialog editor puts it back in sync. Thus, here in 1996, you have some of the biggest pictures being cut by some of the most experienced picture editors and tracks are coming to sound editorial OUT OF SYNC. No longer can sound editorial teams assume the edited worktrack is a reliable sync reference, and this must stop!

•  Use timecode in the field, including timecode slates. Yeah, I know some camera crews piss and moan about the extra trouble it takes to use electronic states, but they will give you (and as a result, us) a definitive way of checking that your production (aka, source or Nagra) timecode for a given frame is the same in the system's database as it is on the picture image.

And, of course, always use non-drop timecode. Shoot on sight anyone who tells you to use drop-frame.

•  Give us printed timelines. CMX-style EDLs are great for telling software how to auto-assemble material but are a pain in the ass to read, not to mention that they're in timecode form, and many of us only think in terms of feet and frames. If you can, please print out a timeline on your system that will show a graphic representation (in both footage and timecode counts) of what scene and take is used at any given point on either Video or Audio 1-4. It's a clear way of showing how a reel is assembled, much in the way an analog watch is better at showing you that you have 11 minutes to go before the hour.

With regard to other paperwork, ask assistant sound editors how they want to receive cut lists and change lists. It's very possible that they will want it on disk, all the better to make custom printouts if they're so inclined (which will save you time, of course).

•  Use good speakers. Don't accept the crappy little white speakers that are rented or sold with many nonlinear edit systems. Courtesy of the world of home theater, there are many high-quality shielded speakers available that will give you a much better clue as to what's on your track.

•  Digitize your picture at the best resolution you can afford. While picture resolution would seem to be out of the domain of sound folks, it becomes a serious issue when you expect us to cut against it for temp dubs. I put a clause in my contracts that specifies quality regarding digital outputs: If it ain't good, there has to be a telecine of a conformed work picture. The primary downside of this is that it lengthens the time between the lock of a new version and the start of sound editing; for temps, a day can be a long time.

•  Talk to the production mixer. Let him or her know that you will crucify them for using EQ before you will crucify them for having too much low end on the track during dailies. Be clear on the instructions for combining 2-track material during the printing of dailies.

•  Document what sound effects you cut into your worktrack. Those of us on the sound end can start the procedure when we first send you effects to cut in. You should never receive from us a DAT that says "gunshot"; it should always be accompanied by a number tracing it either to our master library system (the preferred method), or at least to a film-specific number that is cross-referenced to the original library number.

And you in turn should digitize it under said number so we will know, when cutting sound for temp dubs, what effects we should take from your worktracks and which ones we need to cut ourselves.

•  Lay out your tracks with care. Since it's so easy to cut tracks on nonlinear edit systems, there's a tendency to spread things out, making it hard to sort through everything during dialog editing. A big help would be to have tracks 1 and 2 containing sync production track only. While the dialog editor would want access to both tracks individually to see your edit points exactly, virtually all other uses of the worktrack (such as Foley editing) can use a combine with no compromise; this would not be the case if the crowd walla that you cut for your scratch temps was on track 2 and would muddy everything up.

Please keep music and effects on tracks 3 and 4, and keep them in mono to save track space.

•  Picky technical stuff. Unless you're doing a rough scratch mix that is going to be transferred to mag for a screening, don't ever EQ anything coming out of your system. Place a line-up tone at the beginning of each reel and transfer it to whatever medium you're delivering to us.

Be clear about your signal flow chart when digitizing, editing and stringing off to DAT or MDM [modular digital multitrack]. When you are digitizing a mono track from a DAT, don't combine both tracks during the edit and indeed be very careful about being double-bussed on your console. Too many times I have heard tracks from the picture department that have a phasey haze all over them.

And last, but not least, have nice, fat, clean sync pops at the head and tail of every track.

•  Audio quality. Many of my brothers and sisters in post-production are always complaining about the tracks they get from nonlinear edit systems for temp dubs. This doesn't have to be the case, and one of the solutions is simple: Do not use linear videotape tracks as your sound source!

Certainly, I think, the most common route employed today — production track transferred to mag, which is transferred to Hi-Fi tracks of Betacam or SuperVHS, which is then transferred to the nonlinear edit system's hard drive — is of high enough quality to deal with temp dubs with no embarrassment whatsoever. But because this involves so many scary issues (Did the dailies transfer house properly align the mag machine? What kind of consoles and electronics were used throughout this whole process?), my post budgets always include reprinting the audio prior to final dialog editing. Call me cranky, but I just assume that whoever transferred the dailies was not taking the approach that they were mastering this audio once and forever, that this transfer will extract the information on this tape as accurately as possible. As the supervisor of the soundtrack in post-production, it's our responsibility to get the most out of the tracks that the production sound team turn in; any less, and it's an insult to them.

However, your goal should be to deliver master-quality audio to us so that we will not have to start from scratch by re-transferring and re-editing your production track. In the best of all worlds (which all of us want and many are working on), you can give us a hard drive with your original sound files, and our dialog editing will consist solely of splitting, filling and otherwise finessing what were originally your A and B tracks. While this may not seem very exciting to you, maybe you can look at it another way: This will save producers a fortune in dialog editing and transfers, certainly enough to justify their using a nonlinear edit system if their numbers had indicated that it might be too expensive for the budget.

A large percentage of my worries go away with audio transferred digitally all the way through.

Those are some of the potential pitfalls associated with cutting on nonlinear picture editing systems. It's not that these systems are flawed; rather, it's usually a simple question of everyone learning what parts of each system require diligence. Now, let's focus on how to get the most out of your digital nonlinear edit systems from a more technical point of view — specifically, dealing with some of the more obscure and even less understood aspects of synchronization, which I hope will be more of a wake-up call than a snooze ...

•  Keep in sync. In case you're wondering what I'm talking about here, when I say "sync," I mean that the lips match the sound on videotapes and on your worktrack when "popped." Dead-on in sync throughout the reel up to the tail pop. Obvious, yes, so why am I making such a point of it?

In the course of preparing and editing tracks for the mix, sound editors will frequently need to reference to correct production sync. The most obvious example is the worktrack itself: There is a great amount of guesswork removed from the dialog editing process when reprinting from the original production sound tapes, if we can phase the reprinted material to the worktrack and know that it a) is in sync, and b) accurately reflects decisions that you and the director made during picture editing. Thus, if you have cheated an alternate take, we can be sure that we are honoring your intentions.

When editing Foley, a bottle set-down should be able to be cut in sync by matching modulations with the production track. There should be no thought to whether or not the original take was properly synched in dailies or whether or not the whole track is in sync. Thus, we can be sure that the Foley premix will be in exact sync with the dialog premix, and that both will be in visual "editor gut sync." Flamming of production and Foley is not acceptable.

The production track is being constantly referenced in the case of both Foley and dialog editing, and therefore it is best to have it available at all times, in spite of the hard disk space this eats up. However, this shouldn't be necessary for most sound effects editing, and a door close should be able to be cut to the track on the videotape with confidence. By the same token, an ADR editor should be able to copy the VCR audio track into the edit session only when needed, because they might only have one looped line in a 20 minute reel. They should take for granted that it is in sync with the dialog editor's cut tracks.

Getting our videotapes right means that your picture department is responsible for the sync of the mag dubber during telecine of the workprint. I know I've talked about telecine already, but I need to emphasize that you need to take the same care with the cut workprint as you do with original dailies. The days of handing the sound department reversal film dupes and mag worktrack are gone.

After all this explaining, you're still probably asking: Why is he making such a big deal out of all of this obvious stuff? Because very, very rarely is all of this sync minutiae attended to these days by the picture department, no matter how experienced or otherwise capable. This whole world of nonlinear picture editing has so many people spooked that they use technophobia as an excuse for being sloppy.

•  Know thine start marks. Again, when you were dealing only with film, it was an easy matter to know that the sync pop was 9 feet from the start mark, and the first frame of picture was at 12 feet. And indeed this shouldn't change even though we've thrown in the variables of videotape and timecode. The relationships should stay constant.

You should always regard the first field in which the "picture start" frame in the leader is visible to be the first field of the reel-number-hour timecode (i.e., 02:00:00:00 for reel two), and not one stinking field sooner or later. Ditto for the pop (0n:00:06:00) and first frame of picture (0n:00:08:00). In addition, to keep everything neat, clean and repeatable, all of these video frames should be "A" frames at telecine, i.e., one film frame on one video frame. And don't forget to give us field-accurate burn-ins of both timecode and footage on all videotapes; the larger the burn-ins, the better.

Important note: All of this applies both to videotape copies direct from your edit system ("digital cuts") and to telecine transfers of workprints; if you abide by these rules, then we can start cutting as soon as we get a videotape from you. That is, assuming that the resolution of your digitized image doesn't render lip movements illegible.

•  Cut in 2,000-foot reels. In the days when everything was on sprockets, it was very cumbersome to deal with large reels of film on rewinds. But at this point in the game, it's possible to do a movie with the only sprocketed, edited reel being the picture workprint. Hard disks and videotapes don't know from 10-minute or 20-minute reels; torque, inertia and elbow grease don't apply. This, of course, assumes that the guide worktrack will come directly from your nonlinear edit system, and that mag film will only be used on the re-recording stage.

By turning over reels to the sound department in "AB" projection reels, you will save us (that is, the sound editors and re-recording mixers) time and money by reducing the number of reel changeovers. I estimate that this saves at least a day of mixing time on a simple film, maybe even three when there is a lot of premixing involved. Tack on another day if you're mixing to mag. This will also make matters much easier for us by reducing the possibility of our having to match levels during a reel change in the middle of a scene.

At the very least, should you cut the film in 1,000-foot reels because of massive picture changes, try to join the reels prior to the final mix. (Make sure that we get accurate new videotapes of the conformed print!) The most important reason for big reels is that everyone is able to get a better sense of the dynamics and flow of the soundtrack, something that's evident at the finals when the balance of dialog, music and effects is locked into stone.

•  Create a master 4-track mag film for each reel. Even if the mix will take place in a non-sprocketed multitrack audio/video picture environment, it is always helpful to have a piece of sprocketed mag reflecting your four audio tracks. (At the very least, the material should be on whatever 8-track modular digital multitrack format the mix stage prefers.) Not only will this be useful in reviving the odd favorite effect, but it will settle all sync arguments as to how your dialog or music was originally cut. Speaking of which ...

•  Make sure your music is in sync. The history of film sound is filled with thousands of stories of picture departments cutting to music derived from a nonsync master, only to have it walk out of sync at the mix. (The nadir was perhaps in the mid-'80s, when the arrival of music videos coincided with the beginning of timecode and video in film production.) Again, before you start editing, make tests here just like you did with the original sync test described above. Do a resolved transfer of the music master (I'm speaking here of the God Master, such as a multitrack tape) to mag and then compare this to a mag that has taken a path from the God Master to your 1/4-inch playback master to your nonlinear system (loading in as you would for dailies) to whatever format you're delivering your tracks to sound editorial on (DAT, MDM), and then do a resolved transfer of that tape to mag. The two mags should hold exact phase sync for the length of the cue. If this doesn't happen, call in the propeller heads immediately.

While the above situation implies that you have a timecoded (and therefore synchronous) multitrack master as God, you also should be aware of how to create a God Master when what you think is God is in fact a false prophet. In plain English, if you are taking music from a 2-track medium with no internal sync-pulse (non-sync 1/4-inch, records, CDs, etc.), first make a carefully aligned transfer to Dolby SR-encoded 1/4-inch with timecode (the most universally accepted format, and certainly my first choice) or to DAT/MDM with timecode. I say "carefully aligned" because there will be no looking back at this point, so make sure your source is the best-generation master. To keep everything neat, have the first modulation of each cue start on the timecode hour; it pisses me off to get a timecoded master with the downbeat at 13:29:52:12.

DAT has the advantage that it is possible for you to do the transfers into a nonlinear edit system from a standard non-timecode (read "less expensive") machine. However, you must make sure that you do the dual-mag test described above before signing off.

As I've already suggested, keep music on track 4, going to track 3 when necessary for overlaps or tight transitions. Put music in tracks 1 and 2 only when it is the sync track (live recording or playback), without which there would be no audio. Cut music in mono to save tracks, since it can and will easily be replaced later on.

•  Know your digital audio sample rates and choose them with care. I think it's pretty appalling that the user guides of the nonlinear picture editing systems most popular with feature films have almost nothing written about sampling rates in clear, practical terms. To wit, can you tell me what the sample rate of your system is when referenced to NTSC video? Look in your manual and see if it sheds any light or advice; I doubt it.

Clearly, the long-term goal of those of us in film sound is to replace analog-to-digital and digital-to-analog conversions with digital-to-digital transfers whenever possible. This should, of course, start with the original production recordings (tapes today, who knows what tomorrow) that end up on the hard drives you edit on. One day, it will be standard for sound editors to get those drives from the picture-editing department, allowing us to open up edits without having to re-transfer a take.

Let's concentrate on the picky underlying principles of how production audio can be transferred digitally to the nonlinear edit system. Issues of timecode are purposely left out of this discussion because digital audio sampling rates have a tenuous and bizarre relationship with timecode frame rates and film frame rates: You can mix and match them. The key thing to remember is that you can do D-to-D transfers only as far the relationship of sample rate to camera speed remains constant.

For example, if the production mixer records with a DAT or MDM at 48.000 kHz (the standard rate of most non-timecode digital decks when you simply go into record) relative to 24.00 frames per second in the field, you can do a synchronous digital transfer into your nonlinear edit system only if your sample rate is 47.952 kHz when using an NTSC telecine of your film. Because the film was slowed down 0.1% during telecine, and thus the film speed was 23.97 frames per second, your sample rate at NTSC play speed needs to reflect this change. (The math is similar when working at the 44.1kHz sampling rate: Use 44.1 in the field, slowing down the sample rate to 44.056 when cutting to an NTSC telecine transfer.)

Carrying the equation to my field of post sound, we can accept audio from you digitally only if our sample rates are the same. Therefore, in the equation noted above, if we run our workstations at 48.000 kHz referenced to NTSC color, you can't give us your worktrack on a DAT that was transferred digitally from your edit system, because when we load it into our workstation D-to-D, it will be 0.1% too short since we will be playing back at the original 24 fps/48.000 speed.

The solution to the common problem described above is simple: Output your tracks to a digital deck in the analog mode, running at 48 kHz, simply pressing "record," The track can then be loaded into the sound editor's workstation digitally. Remember to always use consistent levels when outputting in this manner.

The above scenario is very common, and this is why it makes sense to run nonlinear picture and sound editing systems at 47.952 kHz at NTSC speed. An equally solid solution is to run digital recorders in the field at 48.048 kHz/24.00 fps, where nonlinear edit systems would conform to the 48.000 kHz NTSC digital videotape standard. This would be most notably useful when making programs destined for online editing and indeed when doing all postproduction audio for a feature film digitally through a digital console,

Re: working at 44.1 kHz. There is really very little reason to do so since any DAT machine is capable of 48kHz recording (via analog or digital inputs), something that cannot be said of 44.1, which some machines can only record and play back digitally.

In case it needs to be stated, never run your audio at a sample rate less than 44.1 kHz. The hard disk space that you're saving is just not worth the sound quality loss.

•  Be consistent with your version numbers. I have always found it best to change version numbers with each screening or preview, with the first changes that you do after them to a new number. Thus, the day after the version 7 temp, you should do a "save as" V8.0. If you turn over videotapes to us as a specific version, and then you do a change in the track only, increment the version with an alphabetic suffix only — "V8.1a" tells us that the version 8.1 videotape is still good.

As always, ask a lot and assume nothing.

I hope this two-part letter has been of help to picture editors and assistants, and I'd appreciate your sending favorite sound/picture interface stories to: PO Box 24609, New Orleans, LA 70184; fax 504/488-5139; or via e-mail: swelltone@aol.com.

Larry Blake is a sound editor / rerecording mixer who lives in New Orleans for reasons too numerous to mention, although one of them would have to be that it can be so miserably hot there that he can always pull rank when his wimpy friends in L.A. complain about the smog-filled dry air that passes for summer heat.