Fun with H4 Syncing
 

All right - I think I'm a pretty smart guy, but after banging my head against a wall for weeks on this I admit defeat. I simply can't get an H4 wav file to sync up with the video that I shot.

I recorded an interview that's about an hour long. I recorded the sound on both the H4 and the camera's internal mic and imported both into Premiere CS3. I thought that I would just adjust the Speed/Duration of the H4 file and line up the waveforms. The only problem is that no matter what number I use for the Speed/Duration, they never line up throughout the interview.

I've searched for and read tons of information online and tried a preposterous number of things but still can't get it to work.

Does somebody have a method of doing this syncing in Premiere?

- Rick

I don't know Premier's tools. but if you're desperate to fix this particular footage you can download the trial version of Vegas. Its time stretch function works fine with my long (one hour +) H4 files.

there is NO reason to change the speed of the audio. thats why you have a problem. this isn't the days of film where 24fps was telecine'd to NTSC and so became 23.976. if you shot with a video camera, even at 24P, 60 sec on tape = 60 sec on audio HD, nothing more, nothing less.

if the camera or HD recorder dropped out of sync because one of them stopped started, simply use a divide by 2 comparison to find the break point. it works like this if the front is ok, go to the middle and check. if its bad there, go 1/2 from there back and check. if its good, go 1/2 from the current to last check point and check. if its bad, go 1/2 way back. this should require about 7 tries to find the bad spot within a minute or two on a long take.

Without slaving the recorder to the camera (which you can't do with the H4), you can't expect if to stay in sync. However as a past owner of the Zoom crap from (we know where), they are all over the board on clock speed. Hard to do any consistent time stretch....

Very true, But recording sample clocks can be off-spec, especially with consumer gear. Let's say he recorded 60 minutes (3600 seconds) of material with the H4 at a nominal 48kHz but the recorder's clock was really 0.1% slow. One hour's worth of material should have exactly 172,800,000 samples but because of the slow clock it only has 172,627,200 samples. Now that material is loaded into an NLE that is exactly on 48kHz. At the correct sample rate those 172,627,200 samples that were recorded in 60 minutes will play out in 59 minutes 24 seconds. Since the camera can be assumed to be the master, sync lined up at the start of the shot will drift out by 36 seconds at the end with H4 sound leading camera sound.

The OP didn't say what camera he was using but another issue that could be the culprit is the fact that miniDV audio isn't locked to picture and the video and audio sample clocks in the camera itself could be slightly off from each other. They'll be in sync when played from the camera but drift out when played in the NLE. If this happend to him, it could be the reference tracks recorded in camera are slightly off with respect to picture after capture. Trying to lineup the H4 to them would mean he's trying to hit a moving target. I've seen several references that recommend when capturing very long shots one take the video through Firewire but to take the camera audio through an analog interface so they're resampled to avoid the problem of drift between them

There's no substitute for proper slates and for long takes I think getting both a head slate and a tail slate when possible is a really good idea.

I use the technique I outlined here: http://www.mfbb.net/myvideoproblems/...s-about25.html.
to resync externally recorded audio, when the professional options are not available.
Bit long winded the first time, and of course it may be easily adapted to suit whichever audio editor you have, but should only need to be done once for each recorder.

David -

Good idea! I've downloaded Vegas and will give it a try.

Steve Oakley -

There were no dropouts. As Steve House explained, with the H4, 60 seconds on tape doesn't necessarily equal 60 seconds on audio.

Steve House -

I recorded the video using a Canon HV20. The camera was connected to a Blackmagic Intensity Pro via HDMI and I recorded the video uncompressed. The audio that was was recorded via the camera is synced up with the video - no problem there. I also did slate the take but even with the slates I'm still having problems.

Roger -

I did find and try your technique, but couldn't get Audacity to work. It kept crashing whenever I loaded one of the sound files. Other programs, like Premiere, load the files but Audacity just doesn't like them. I'll be trying some other audio programs as well.

Thanks for the suggestions - I'll try them. Hopefully I'll get this thing to work!

- Rick

all I can say is starting in the 80's shooting 16 & a Nagra, this wasn't a problem.there was certainly no cable between camera and tape. yes you did have a slight speed change due to telecine, but that was a known factor once done, all was good. If the H4 doesn't have a stable on time clock, why use it ? I'd make who ever sold this to you take it back as defective product.

a couple of years ago I ran a DVCpro 450 deck to record a board feed from a show basically using it as a 2hr 48k 16bit recorder. ran 4hrs thru the deck. locked seemlessly to the 3 cams shooting. deck was self referenced. cameras where 300ft away from the recorder with no connection of any sort. there are plenty of accurately timed recorders out there. I was interested in one of these low end recorders, but if they don't have good clocks, it really defeats the purpose of using it and being untethered from the camera.

in your case there is but one somewhat easy fix. play thru it and every time it starts to drift noticibly, cut the audio track ( ideally between sentences or pauses, patch gap with room tone ) and move it + or - a frame so it locks again. repeat until end.I've seen this done many times. if the h4 varies its clock, this is the only fix. if you have access to pro tools and the ADR pluging they make for this, it would pretty much be automated.

Audacity
 

I find Audacity rock steady and very easy to use/very useful - it will be ideal in what you're trying to do (I have a Zoom H2 so know the issue you're talking about well!). Did you download the new beta version maybe? - if so I suggest you try the earlier version (both on this link.)

Also, just seen Steve's reply and yes I sometimes do that too - cutting the sound file up into smaller, easier to manage segments works well in Vegas 7e and Ulead VS10+. I find syncing audio this way VERY easy to do.

http://audacity.sourceforge.net/

Clock stability is certainly one of the things that accounts for the price difference between a ~$400 recorder like the Zoom and a ~$2500 professional recorder like the SD702/702t. But even Sound Devices cautions that drift may occur on shots running over about 15 minutes in duration because some cameras might not be spot on.

I don't find the H4's clock to be unstable, it just runs fast resulting in an audio track that's too long. So I shrink it. A lot has been posted about it.

Problem is, once you download the video from your camera into the NLE, then it (and it's onboard 'sync' audio) become the master track, so even if your remote audio recorder has a more accurate clock than your camera, it's the recorder track length you'd have to change!

Yep, you're exactly right.

Strictly speaking I shouldn't have used the term "unstable." "Inaccurate" would be more correct. To hold sync, the camera clock and the recorder clock must be exactly the same, whether that's exactly 48kHz or not.

How about cutting the audio file into shorter segments, and line them each up separately? You should not have so much drift in, say, 10 minutes.
Patrick

Don't adjust an H4 or H2 audio track in segments.
 

I have an H2 (not an H4--I misspoke above). It's 16 frames too long after an hour. That would be a highly unacceptable 2-3 frames after 10 minutes. So if I cut my track and shifted it, it would be in synch in the middle and still an unacceptable 1-2 frames off at the beginning and at the end of each segment.

So don't cut up your track--that will only make more work and give a crappy result compared to simply shrinking the whole track. Shrinking the track gives virtually perfect synch (within an unnoticeable fraction of a frame). Done.

If your program doesn't do stretching and shrinking, then maybe segmenting is your only choice. But you can get another program.

Rick, did you get a chance to try the Vegas demo to fix your track?

All -

I tried Roger's procedure using Audacity and it worked! I still had the occasional crash with Audacity but had it going enough that I was able to complete the procedure (Andy, I am using the earlier version - for some reason it just crashes occasionally).

I tried the procedure with a 33 minute clip and got my magic Change Tempo - Percent Change number of 0.009. I fixed up the audio and it looks good so far. Right now I'm in the middle of processing a longer clip - crossing my fingers!

David, it looks like the stretching and shrinking procedure is going to work so I didn't try the Vegas demo. Thanks for the suggestion, though!

I was thinking... How about next time I feed the line output from the H4 into the mic input on my HV20 and record the audio with the video track? That way I can just skip all this syncing silliness and have a good audio track ready to go right away!

- Rick

Shrinking is the process that I recommended the Vegas demo for. If you got it another way, great.

If you're going to do that, why bother with the H4 at all? A mic preamp or mixer would be better suited. One of the reasons for going double system is to record the audio at a higher quality than possible with the in-camera audio circuitry. If you're going to ultimatly use it anyway, simplify and cut out the middle-man.

Steve -

The main reason I went for the H4 was because I understood it to be a way to get good sound at a reasonable price. Getting a mic with the associated equipment is another fairly large expense that I want to avoid if possible.

I was just thinking that I could take the signal from the H4 and instead of recording it to the SD card, I could send it to the camera and have the camera record it along with the video feed. Would the audio be degraded by having it saved that way? If not, it would save me the bother of having to sync it afterwards and I wouldn't have to buy any additional equipment.

- Rick

I can't comment on the relative quality of audio tracks recorded in the HV20 versus the H4 as I don't have hands-on experience with either one. I can only say that in general the audio on consumer cameras isn't noted for being the best in the world and your results are only going to be as good as the weakest link, which in this case I would expect to be the camera audio. So feeding a mic to the H4 but not recording it there, instead sending the signal on to the camera to record with the video would probably not give you any better results than plugging the mic directly into the camera. The H4 would be relegated to the role of an external preamp at best and any deficiencies in the camera audio would still be in play. Another issue to be aware of is the need to match levels - I believe the HV20 only has a mic-level input while you'd be coming out of the H4 at line level and its output would be much too hot for the camera input. Overloading the camera's inputs could actually make it sound worse.

I just recorded an hour and 7 minute presentation to sync with the video. My H4, just purchased from BH, is fast by about 1.87 seconds over that time frame.

It's a hassle to cut and resync, but it does have the XLR's which is handy and the quality of the recording good. Meaning no hum, clicks, etc. Although, the recording level was very low. I recorded from a line out of the wireless station using 1/8 plug and adapted to my two xlrs. Very low on the meters and I had to amplify significantly in Audition.

It's doable for the price, I suppose.

CORRECTION: The H4 recording is SLOWER by 1.87 seconds, not faster.

Also, shrinking works perfectly instead of cut and sync. I went from 100 to 100.05 and it lined up perfectly.

I've got to agree with you there Bill - although I linked to a procedure I devised using free programs, I find it even easier to do, as you do, using Audition (or in my case, CoolEditPro2).
I firmly believe the the audio sync problem is nearly always a result of the difference in device clock frequencies, rather than the 'drift' of any individual device.
It may well be that the camera clock is the one that's at fault, but as it is the one that is intrinsically linked to the audio, then the remote recording device has the audio that has to be shrunk (or stretched!). And if you know in advance by how much, then it really becomes a much simpler task.
Definitely easier than doing it in small 'chunks', IMHO.

Of course, we should all really use synced timecode, but in the real world, where budgets are sometimes tight?......

Just FYI - it's a common belief that synced timecode alone will keep the two devices in sync but it's not correct. Timecode is like a slate and recording identical code with the video and the audio lets you pick any arbitrary point in the video timeline and align the matching point in the audio timeline to it. But timecode doesn't control the recording or playback SPEED of either recording and if their clocks aren't running at exactly the same rate when making the the recording their lengths will be different - if you align the codes at the start of the file you'll find they're out of alignment at the end, if you line up in the middle of the file you'll both the head and tail will be off, etc. With film and analog recording or a DAT that's going through telecine it's another story because the timecode recorded on the audio tape or file controlls the speed of the playback device that resolved magnetic tape to perforated magnetic film or the speed of the telecine transfer. Sound for video with file-based digital recording is a whole different process. To get the video and audio to stay in "sync" over the duration of a shot, the video sample clock in the camera and the audio sample clock in the recorder must be slaved one to the other or to a common source clock, using genlock I/O on the camera with wordclock I/O on the recorder or deriving wordclock for the audio from the video blackburst on the camera. That makes sure that the audio and video files are exactly the same length when loaded into the NLE. Then you can couple that with timecode to aid in establishing the initial lineup of the two files and you're in business.

The way big0time multicam shoots manage it is they have a master "house clock" that generates timecode, video sync, and wordclock. TC plus video sync goes to all the cameras and TC plus wordclock goes to the audio gear. That's where "Lockit boxes" come in - they're tuned to the master clock and then attached to the various devices to provide identical clocks to all of them without requiring hard-wired cabling. Now everything has common TC and a common clock reference.

I've never quite understood the fuss about timecode synching for audio with video. If you have a very large number of tracks, I can understand the desire to avoid having to make a lot of adjustments in post.

But when I have three camera tracks from a multicam shoot, plus a few additional imported audio tracks (house mixer, my own orchestra pit recording, etc.), what's the big deal? It's going to be an 80+ hour editing project anyway, so a half-hour or so of that is aligning audio tracks. Not bad.

The resulting synch is as good as I think I see and hear at the movies or on TV. Am I wrong? So other than the occasional studio pro, what are people on this board doing that makes them obsess about this?