Audio with dslr cameras

Evening all and happy long weekend coming up!
(well, for those of you with jobs anyway)

A few of my friends recently entered one of those 48hr film contests. As the name implies, they were given a topic and had 48 hours to write, shoot, edit and burn a dvd of 7-10 minutes in length.
They chose to use a dslr camera, and to record the audio separately instead of using the camera mic (or maybe in addition to it).

What is the proper way to marry the audio with the video in this type of setup. I wish I could give more details, but I wasn't there. All I know is they were using FCP7.
Apparently the video looked great, the audio sounded great ... but the editor couldn't make it work and they missed the deadline for submission.

Can someone point me to a tutorial of do's and don'ts for working like this?
I may be using this technique myself in the near future and I would like to know about any pitfalls before I try it.
My limited understanding assumes you would drop in the video (uncut), then lay the audio in like a soundtrack and sync up with the little clapboard thingie. Once it is synced up, can't you cut them "together" and move them around?
The kicker here is that the editor actually teaches a college class on editing (FCP). Makes me wonder if this is really a lot more complicated than I think it is.

Thanks,
Rick

> They chose to use a dslr camera, and to record the audio separately instead of using
> the camera mic (or maybe in addition to it).

Tell them right away: Do not, do not, do not forego the camera audio. Always leave it on. This kind of non-professional dual-system sound has a high failure rate. The camera sound could end up being the only sound you have on some or even all of the takes.

In fact, on low-budget shoots, you'd want your camera operator, an AC or the director/producer (anybody close to the camera) to be listening in on the camera audio. Ideally your camera would have manual audio gain controls. Usually the problem is that the camera audio is too loud, not too soft. So, always set it on the low-ish side and make sure the audio doesn't go into the red.

> What is the proper way to marry the audio with the video in this type of setup.

Make sure both camera and sound recorder are set to 48kHz. Make sure you do verbal slates; low-end audio recorders do not have the option of choosing your file name, and a verbal slate -- "Scene One, Shot One, Take Six" -- will drastically shorten your sync time since you can identify every audio take by listening to the first 10 seconds of it.

Do not cut corners by recording every take of every shot, or even every setup, as one long audio take. The few seconds you save on set will be surpassed by the extra time needed to organize, break up, and sync the long piece of audio later. Also, one long piece will multiply the danger of the recorder failing after 30 minutes of recording, and you end up losing all 30 minutes because the recorder doesn't write the file until you stop. (Newer recorders like the Zoom H2n supposedly have backup features so that a sudden failure doesn't erase the file, but preserves it up to the point of failure. I can't vouch for this feature, however, as I haven't yet used a recorder that has it.)

Sync video and audio clips in a Sequence. Use markers to mark the sync point (eg. clapboard). When you sync, make sure your sync markers are made both in the affiliate clip (the one in the sync sequence) as well as the master clip in the Browser.

Preserve the sync sequence in perpetuity; don't ever delete it. Then link the video to the audio and make a new master clip by dragging the linked pair to a bin. Plan on doubling your editing time for this operation. PluralEyes can use camera audio to locate the correct audio clip and sync it automatically. I haven't used it, but I don't think it can completely replace human experience. It can help with the first stage, but you'll still need to know how to work with the clips afterwards.

My suggestion would be to keep very good notes (see verbal slating above), do your assistant-editor work thoroughly (eg. strict, disciplined organization of both video and audio footage before importing and editing), and to try to edit the piece before dealing with sync. That way you're syncing 1/20th of the shot footage, rather than all the dailies. Usually you'd want to sync first, but that will take a lot of your time. If you organized the clips well enough, you should be able to just sync the footage that is used in the final picture.

> Can someone point me to a tutorial of do's and don'ts for working like this?

In my opinion, a tutorial won't be very useful here. You can follow a lesson, but preserving sync is one of those long-term editor's skillsets that can only come of having dealt with it on project after project. It also requires considerable hands-on knowledge of the actual editing software in question; preserving sync in Avid, for example, is completely different from preserving sync in FCP. Unless you've already done it yourself, hands on, reading about it on the web won't really sink it into your head and your editing habits. Just do it over and over, on multiple projects, and learn from your mistakes. We all have.

> the editor couldn't make it work and they missed the deadline for submission.
> The kicker here is that the editor actually teaches a college class on editing (FCP).

There isn't enough information to know whether the editor was weak on technique or whether the director/producer/crew screwed it up. If they'd recorded, say, all 24 shots of Scene 01 as one long chunk of audio, didn't do any verbal slates, the camera audio was off, and the audio was recorded at 32kHz so that it ran at a different speed from the camera, then I might not have been able to sync it up in time, either. That kind of mess-up also makes it nearly impossible to edit first, sync later, since hunting for the correct take would have taken a long time. People using video cameras are also often incredibly sloppy with visual takes -- eg. doing too many "series" shots, leaving the camera on when they should just shut it off and restart. That also drastically increases difficulty in syncing.

Did they actually consult the editor before proceeding with the shoot? If they didn't, they deserved to miss their deadline; they didn't do their homework.


www.derekmok.com

Hi Derek,
Thanks for your input. Just what I was looking for.
Sorry I don't have all the particulars. When I find out more, I will post them.


Thanks again,
Rick

ah... an interesting and timely topic for me.
i am syncing rushes, or rather testing some workflows

but first i have to respond to what Derek said:
"try to edit the piece before dealing with sync. That way you're syncing 1/20th of the shot footage, rather than all the dailies. Usually you'd want to sync first, but that will take a lot of your time. If you organized the clips well enough, you should be able to just sync the footage that is used in the final picture."

i have to say i completely disagree with my learned colleague Derek on this!
that way will just get you into trouble, and may well be impossible on many types of program.

how many times do people post here saying, i cut my show with the camera audio, so how do i sync it to the good stuff, now that i have fifteen hundred cuts in my show?
too many times, especially since the answer is always: "it's really hard and you should have done your syncing first."

even if D is just talking about a rough assembly, i'm very much of the school where i do want to have everything at my fingertips.
what may have seemed like an NG on set, could be gold.

i know that it's much more efficient to do all that assistant work (syncing, linking, sub-clipping) first, then dive into the editing.


as for the practice of syncing rushes, there are may different approaches.
if you DONT have matching TC on picture and sound, then this is what has been working for me:


......................................................................................

lay all pic into timeline
find slate on first clip, add marker to clip
load audio into viewer, find slate, add marker, mark in, edit.

in timeline, tidy up the end overhand of audio,
drag start of audio back to start of pic.

go to the marker, check sync on the slate,
go into the clip, closer to the end than the beginning, find some sync, and check it.

all ok? back to the slate, NAME the clip**.
move onto the next slate.

......................................................................................


as people around here know, i use the macros app quickeys.
i LOVE quickeys, and you would do after using it to sync rushes with my macros!

all of the above is done with 2 macros.
the human involvement is ONLY:
finding the pic slate,
finding the audio slate
checking sync
naming the clip
everything else is automated.... it is very quick!
(and if the sound recordist has been labelling the files correctly, then naming the slate can be automated, too)

however, for a one day rush job, it might not be efficient to work out and create new macros. (certainly if you've never done it before) that does take some time.

anyway, after i get to the end of the timeline, i go back and link each clip with it's audio (select picture and audio, Apple L)
then open it into the viewer (return)
and make a subclip (Apple U)
(then remove the word "subclip"

this trims the audio to the length of the picture. much neater.

of course i have a macro for that, too.

no, hang on...
there has been a bug with FCP7 where this process eventually bogs FCP down with weird behaviour.
our new workaround is to link the clip,
copy it into a in in the browser,
colour it
open it from there into the viewer, then subclip
remove the colour from the subclip.

when all is done, we delete the coloured clips, which are not subclips.
it's more steps, but with a macro you wouldn't notice.

after the syncing, linking and sub-clipping, we move onto creating scene bins in a separate project.
oh, yes, a project for syncing, a project for scene bins,
and new ones each day.


tonight i have been experimenting with media that has matching TC, using Spherico FIlm-Tools "Sequence Liner" to line up the picture and sound.
you would think that it is quicker to deal with, but i'm not so sure.

after all, you wouldn't entirely trust that the TC will perfectly match, so you still need to check,
you need to remove gaps, and generally tidy up the rushes,
and to add the name of the clip to the picture.

i've been build a macro to deal with all of this (and in hits particular case the audio has a version of the slate name, so i can copy / paste / modify it)
but still i don't think it's is much faster than the more traditional way i outlined above.


if you were using Pluraleyes, well that would take even longer than Sequence Liner.
SL matches number, Pluraleyes has to "listen to" and analyse the audio, which in my (limited) experience takes quite some time.
then of course the DSLR audio is always out, so you have to check it all.

if the shoot went well, and the steps outline by Derek were followed (individual files for each slate, picture & sound; clear calling of the slate with a clear visual slate) then i'm pretty sure i can sync faster than Pluraleyes,
IF the syncing is being done in a methodical manor.


but really, double system on a 48 hr turnaround?
who where they kidding?
that was a very bad idea.
i'll be the filmmakers who made it on time used single system.



nick




**yes, NAME the clip!
many here will disagree, but that is what i have been doing lately.
i can't edit with meaningless file-name gobledy-gook as my clip names.
i need slate and take.
we place the file name into a master comment column prior to syncing.
(using a macro)

nm

>> "try to edit the piece before dealing with sync. That way you're syncing 1/20th of the
>> shot footage, rather than all the dailies. Usually you'd want to sync first, but that will
>> take a lot of your time. If you organized the clips well enough, you should be able to
>> just sync the footage that is used in the final picture."
> i have to say i completely disagree with my learned colleague Derek on this!

You missed part of my post, Nick:

> Usually you'd want to sync first

On most projects, yeah, sync everything and don't start editing until everything's in place. But on a 48-hour film contest, depending on how much footage you have, that could be prohibitively time-consuming. Especially if the crew didn't prep properly.

On such a film shoot, you have to consult the editor. You aim to get usable camera audio, even if it's not good. Because if you miss the deadline, it doesn't matter if you have perfect sync and followed the proper workflow; you're out of the competition. I've seen winners of these contests with audio that sounded like exactly that -- camera audio. But the films worked.

> i'm very much of the school where i do want to have everything at my fingertips.

Same thing. On every project but a 48-hour film contest, you'd want to log and perhaps even capture/transfer all the takes, even NGs, if only to know where the takes are, and know that there isn't some "magic take" in the director/producer's head that was missed. But on a 48-hour film contest, a good 16 to 30 of those hours are spent on writing and shooting, and then you have to budget time for post-production deliverables. On such a project, I'd import/capture only circled takes when available, as long as there isn't, say, just one circled take out of 24 takes; in those cases, I would capture/import at least one or two more alternates.


www.derekmok.com

Sorry, Derek, i did miss that,
apologies.

nm

I've even done a 36-hour film competition. I think it was called Instant Films. And the film I was doing was a long, six-minute-plus dialogue scene. Dialogue scenes are a lose-lose situation: If you sync first, it takes a long time. If you don't sync first, trying to sync 100 snippets that are cross-cutting from Take 1 to Take 4 to Take 2 after the edit...ouch. And this was before tapeless really caught on, which means we had DV tapes that had to be logged and captured.

Again, consult the editor first before any such shoot. In the above scenario, if I had been asked beforehand, my advice would have been: Mix the audio into the camera. Don't use dual-system sound -- or, use the independent sound source as a backup to the camera audio, not the other way around.


www.derekmok.com

"If you sync first, it takes a long time"

not with macros!
it shouldn't take that long to sync,
but obvisouly, it's quicker if you don't have to


nick

That is awesome, Nick.... QK forever.

- Loren

Today's FCP 7 keytip:
Summon your Video Scopes with Option - 9 !

Your Final Cut Studio KeyGuide? Power Pack
with FCP7 KeyGuide --
now available at KeyGuide Central.
www.neotrondesign.com

well, putting LTC from the sound recorder onto one of the sound tracks of the cam gives you identical tc.
With my software you can sync a couple of hundred Bwfs and QTs in less than a minute.

Syncing by hand is possible, but only if there is a decent clap. If there is not, it is hell, especially if you don't have any reference.

I'm currently working on a way that uses the cam's internal clock to sync. This way, you only have to sync the first (and perhaps last) clip by hand, and the rest will be done for you.

Bouke
www.videotoolshed.com

>I'm currently working on a way that uses the cam's internal clock to sync.

Hi Bouke,

That's an interesting concept, however I have noticed that on some cameras, the internal timecode drifts whenever there is a start/stop on the camera when the camera is set to record internal ToD TC without an external sync device.



www.strypesinpost.com

Well, it's even worse.
ALL devices drift. (My mac is slow a couple of seconds over 24 hour...)

That's why i wrote: > (and perhaps last)

The logic behind that: (pointy ears mode ON)
All device clocks run slower or faster than the 'real clock'. But, they all are stable in their speed.
Thus, if you sync the first and the last, you know (well, my soon to be ready software knows) both speeds, so the new sync points can be calculated accurately based on this speed difference.

In theory, you only have to sync two clips once for two given devices (read, cam and bwf recorder), and save the speed difference.
After that, you only need to sync one video to one audio clip. The speed difference will be the same, so the math can still be performed.

I've tested this under lab circumstances, and it works.

If you have understood this, you should be able to see the logic in the following:
It does not matter if you sync the first and last clips, you can sync any two, as long as the time difference between them is large enough to do accurate calculation on the difference.
An hour or so should do the trick, but more is better...

The good part: if this works, you only need one clap per time one of the clocks had a reset (but do it a couple of times a day anyways...)
The bad part: you still need a BWF recorder. The ZOOM recorders ' should ' be BWF recorders , but i have not seen clips (yet) that actually have a meaningful timestamp. And the file creation time is rounded to whole seconds, so that's no help...
(well, in the case from the OP, that would have brough the editor as close as one second, should be doable after that...)

Now, it is possible to turn your average el-cheapo Record-0-matic solid state recorder into a piece of gear that is actually usable, but that is something for another time.


Makes sense?

Bouke
www.videotoolshed.com

Well, aside from that drift which is natural, what I suspect is that latency on entering record mode may affect the internal clock in the camera. This may make the calculation a bit trickier, because then you'll need to factor in the lag time on each camera (provided the lag time on a camera is constant) and the number of times the camera starts/stops.

If all cameras start and stop rolling at the same time or if they shoot the same number of clips over a period of an hour, there would probably be less drift. In fact, I was working on a show, where I had camera guys running and gunning without an external sync device. Sync was off arbitrarily. Some by up to a few seconds, although they have been jam locked at the start of shoot. So I told them to go for longer takes in a multicam scenario simply because there would be less clips to manually sync up. What we also discovered was that when that happened, the sync is off by much lesser amounts.

In other words, if 3 cameras are jam synced at the start of the shoot, and Cam A shoots some 10-15 clips of uneven breaks and durations in an hour, while Cam B shoots one clip for the duration of the hour, and Cam C shoots 3-5 clips for that hour, Cam A may lag drastically more than Cam B, while Cam C will lag a little behind Cam A.



www.strypesinpost.com

Oh dear, you just made things more complicated!
(but still doable, if indeed the start/stop hit offsets the clock with a fixed amount...)

What cams did have this behaviour?
(there is no logic in it, unless the clock is somehow related to the recording part of the system, and i can't understand why that would be the case if you run free run instead of rec run. So i suspect your cams are able to do rec run as well (a very educated guess, not only by this logic, but also cause you can jam them :-) )

And, how bad was it?
(If it's just a small amount, it does not matter if the amount is fixed or not, an average will be sufficient.)

Bouke
www.videotoolshed.com

I was working with the Panny HPX 502s. But I should examine the timecodes again...



www.strypesinpost.com

Haven't seen any mention of Plural Eyes?

Didn't read my post?


www.derekmok.com

Plural Eyes helped me sync an entire 5D show. I'd say it got about 85% spot on. The others were either MOS, or the in camera audio gain was too high and distorted there was nothing it could match too or the slate was too far away from the camera. It wasn't exactly fast, but I was also using "try really hard" and "level audio" because in camera was horrendously loud---- those options though definitely add to the time it took.

What I would do is put all my clips in a timeline chronologically, throw all my audio in same timeline chronologically (this helps speed it up) Any place that had long roll out I would slide to get it a bit closer (probably a bit unneccessary). Options wise in plural eyes I'd go single output sequence, replace audio. By looking at the timelines Plural Eyes created I'd know which ones didn't sync properly. Generally my timelines were full days (sometimes two days)- 2-4 hours long. I seem to remember it being about real time with my options.

I could manually sync faster, but all this happens in the BG also, so you can continue cutting while it's working. Then again, for a day long shoot might not be much help. Sometimes the old fashioned way is the best way.