Filmmaking

Avid Media Composer is not (always) frozen

Tuesday, November 17th, 2026

Avid Media Composer is a complicated application that does all sorts of complicated things, and its user interface only surfaces and explains some of its functions to the user as they happen. Sometimes a background task will take minutes or even hours to complete, and the user interface will hang while waiting for a completion message from a background task, and then come back to life once that task finishes. One way to tell whether Media Composer is actually hung and should be force-quit, or if it’s actually busy and you should wait, is to spy on its filesystem activity. On macOS, I’ll usually run this command in a large Terminal window off to the side, just to be able to see what Media Composer is up to:

sudo opensnoop -v -n AvidMediaComposer

If Media Composer is alive and working, you’ll see messages scrolling by that indicate files and folders the application is accessing. Anyone walking by your edit bay will see this and think you’re hacking the Matrix.

So for example, if you’ve opened a large collaborative project after a vacation, and Media Composer sits there with the message “configuring audio waveform cache” displayed on screen for minutes (or hours), and maybe your mouse pointer has even turned into that spinning color wheel animation that bodes ill — run that command and watch what Media Composer is doing. In the case of that “hang”, you’ll see in the output of that command that Media Composer is not sitting idle but is scanning every bin and folder in the project repeatedly.

In the case of that particular situation, you’ve just gotta let Media Composer continue what it’s doing. If you force-quit, remove the MC State, or all the other normal steps I’ve tried, Media Composer will go right back into that same state when relaunched. It just has to finish whatever it’s doing, which typically takes from 5 minutes to 3 hours.

That’s the TLDR;, the short summary. I wrote up a much more detailed and speculative version of the same in an email a while back, and am going to excerpt it here in case any of the information in it is of use to anyone.

The very very long version from an email

Please note: This lengthy writeup may contain some information that is speculative or inaccurate. I reserve the right to be wrong.

I have good news and bad news about the “configuring audio waveform cache” bug. The good news is that after years of fighting this bug I’ve figured some things out. The bad news is that I’ve found no satisfying workaround and the best thing in most cases seems to be to plow through and let Media Composer finish what it’s doing when it sits there on that message. But that’s an oversimplification and I’ll give you the more involved explanation which includes a note on what can bring on the bug and even a terrible workaround.

Two assertions about this bug that I’ll open with to court controversy:

• The bug is not exclusively related to the use of Mimiq (I’ve seen it on bog-standard machines just connected to a Nexis).

• And most counterintuitively — it has nothing to do with the audio waveform cache. I enjoyed being informed of this one recently after spending way too much time over the years trying to figure out workarounds related to that cache.

My current understanding of the issue starts with how Media Composer opens a project. I’ll start there.

The Project .avp file

A project on a local drive has a project_name.avp file at the root of the project directory, and this gets opened when the project is opened. Nice and simple.

In a collaborative project opened on a Nexis however, this single .avp file won’t do, as there might be 10 or 15 users all trying to open and write to that same file all at once –chaotically clobbering whatever project settings and caches are stored in that .avp file. Instead of everyone opening the same .avp file, each user gets their own folder at the root level of the project workspace named after their computer’s name, and inside this folder is stored that user’s .avp file, audio waveform cache, MCState, etc. 

I heard an Avid Tech call these folders “user bays” once. I don’t know if that’s the official term but I’m going with it. So for example the machine Sharon-WFH gets a user bay folder named “Sharon-WFH” at the root level of the project workspace, and inside it you’ll find Robocop5.avp. In this example, every user bay contains its own Robocop5.avp. There was a time that I thought Media Composer simply copied the main .avp file into each user bay, but that would be too easy. They’re not copies, they’re individually built, and that building is the pain we’re here to discuss.

The slowdown

When a user opens the project, Media Composer does a bunch of maintenance tasks, including updating the audio waveform cache (the source of that progress bar). After the waveform cache is finished, it starts updating or building that user-specific .avp file. If the project has been opened fairly recently, this process of updating the .avp file is so fast that the delay isn’t very noticeable. But if the project has a ton of folders and bins and hasn’t been opened for a long time (or ever) on that computer, the delay can be considerable. It can appear as if the application has frozen, and I’ve seen it take 5-180 minutes to finish this task. All the while, the “configuring audio waveform cache” message is displayed –not because the audio waveform cache is being configured*, but because that task’s message isn’t cleared before starting the subsequent, .avp building step in the process of opening the Project. 

A fun Terminal command to spy on Media Composer

One thing I always like to do when Media Composer is stuck in this frozen-looking state is to run this command in a macOS Terminal, to spy on Media Composer’s filesystem activity. It’s interesting and is also a way to be reassured that the application is indeed doing stuff and is not actually frozen (even though it shows up as “non-responsive” in activity monitor). It’s been helpful for this specific bug.

sudo opensnoop -v -n AvidMediaComposer

What is Media Composer doing during that delay?

Beats me! What I can see it doing when I run that Terminal command is that it seems to be scanning through every folder and bin in the project, and then it scans through them again, and again, and eventually it finishes up and opens the project. My guess is that the .avp file contains some sort of representation related to every bin in the project. It’s gotta be a fairly small representation though, as a .avp file for an 800+Gb project might be under 40Mb. I’d love to know why it has to scan through all those bins and why it does so multiple times.

When is a delay most likely?

As previously stated, one is most likely to run into this problem when opening a project that has a lot of bins and folders, and if it’s been a long time since the project was last opened on that computer. This can happen on machines connected to any shared storage that are running Media Composer with collaborative features. In fact we ran into it today opening a large project for the first time on a couple of our computers connected to a Nexis, with no Mimiq in sight. 

This also occasionally seems to occur seemingly at random every few months on every one of our machines, Nexis or Mimiq. I think Media Composer must decide on occasion to do a more full refresh of the user-specific .avp file, but I don’t know its pattern.

When might a delay be more likely on a work-from-home system?

I would imagine that if this scanning were to occur on a project accessed via LucidLink storage, that it would likely take much longer to scan and build the .avp file than if the project were on local media (or locally-shared SMB shares) or a Nexis, depending on internet speed and how LucidLink’s caching is configured. It could even be that the scan takes enough extra time when files are accessed via this current method that the delay wouldn’t be noticeable on other storage. But I’m just speculating as I have no actual LucidLink experience. There may be some Goldilocks range of project sizes at which the user won’t notice the same scanning process delay happening on local storage but would very much notice it on a LucidLink drive. 

There’s also other unrelated weird things that can happen on home networks that can trigger this problem. This might be a rare case but I’ll relate it just in case you run into something like it.

One of our editors (name changed to protect the innocent) kept running into very long “configuring audio waveform cache” delays every few days, even though she’d clearly recently opened the project. She’d still have to wait 90 minutes or so before the project would open. It was only happening to her, and only on her WFH machine. After a week or two of trying to catch this problem as it happened, I finally noticed that she didn’t just have one user bay folder, she had 5 or 6 of them. Instead of just a Sharon-WFH folder at the root level of the project, there was a “Sharon-WFH (2)”, “Sharon-WFH (3)”, on up to “Sharon-WFH (5)”. Each of these folders was its own user bay containing its own .avp file, and each represented her having to wait a very long time while that particular .avp file was compiled from scratch. The reason this happened is that her computer name had changed multiple times.

My guess is that her computer would crash or she’d reboot, and her home router would hold onto the DHCP reservation for her prior computer name for too long and the name would be unavailable. For example her computer would be named Sharon-WFH, and then after a reboot macOS would probably greet her with the message ‘The name of your computer “Sharon-WFH” is already in use on this network. The name has been changed to “Sharon-WFH (2)” automatically’. She’d hit OK on this and think little of it, launch Media Composer, and when she’d open the project Media Composer would see no user bay folder named “Sharon-WFH (2)” and would get started building one and doing its long process of scanning everything to build that user-specific .avp file. 

I haven’t seen this computer name issue show up on our work network, so I speculate that it might be related to particular home router equipment and thus is probably more likely to be correlated with Mimiq use even though it’s not itself actually caused by Mimiq.

Failed workaround attempts (many from before I had much understanding of the bug)

• Force-quit and relaunch media composer – it goes right back to the same delay when next opening the project.

• Rename the user bay –  back to the same delay. 

• Delete MCState and Settings – no effect.

• Copy over searchDBs from someone else’s user folder – back to the same delay. 

• Turn off or disable phonetic indexing – not a bad idea since we aren’t using it and it’s cpu intensive, but it had no effect on this.

The path not taken

• I did once suggest that if we had a smaller project with fewer bins this delay would be reduced, but our team finds it useful to work in a more comprehensive project and it hasn’t cause other problems. 

Partial workarounds

• If needing to just access a bin, perhaps for a screening that can’t wait for that whole .avp building delay, one can make and open a small or new project rather than the main project. That’ll open quickly, and then just open the specific needed bin(s) from the Finder. One doesn’t have a main project open to open bins that live within that project — and locks still work.

• This one didn’t fool media composer for long, but allowed our editor to get into the main project quickly. Here’s my documentation (paraphrased from my Avid forum post) of this convoluted partial workaround procedure just in case a desperate measure is ever required:

1. I made a copy of the project on another workspace

2. Deleted almost every bin and folder in it so there would be less to scan

3. Deleted the user bay folder 

4. Deleted the computer name folder in that project’s AvidSharedData

5. And aimed Media Composer at that project instead of the real one. 

It spun up, recreated those folders, did its ‘configuring waveform cache…’ delay but scanned a much smaller number of bins and folders multiple times before finishing, and then it opened the project. I copied the newly-created user bay folders back into their spots in the original project and aimed media composer at it and lo and behold it opened and seems to be working.  

This workaround has limitations. The delay was back within a day or two when Media Composer realized it needed to update the .avp file. I don’t know if this process will always fool Media Composer as I don’t know how often it checks or how it determines that a .avp file is out of step with a full project. And it’s possible that using a user-specific .avp file that was created from a smaller copy of a larger project will be missing some information that’s actually important (it’d really help to know just what’s in those .avp files). The editor didn’t report running into any issues.

The real solution

The real solution, or what we’ve settled on, is that we let Media Composer do its thing and finish whatever it feels it must finish. None of the workarounds we’ve found stick, and if we don’t let Media Composer build that user’s .avp file to its satisfaction it always notices and decides to pick right back up and start building it again. But we now know to anticipate this delay and build that into our plans. For example, we have a screening tomorrow and will be using a machine on the Nexis that hasn’t previously opened it, so we opened the project on that machine a day in advance, expecting that it would be stuck at ‘configuring audio waveform cache’ for a couple of hours – and that did happen. But now that it’s gotten through building its .avp file, it’ll open the project without delay tomorrow and probably uneventfully for the next couple of months.

A workaround to keep Apple Compressor from transforming DnxHD quicktimes into dark-red freeze frames

Saturday, June 30th, 2018

If an DNxHD quicktime file (reference or self contained) contains only a single video track, it will turn into a dark-red freeze-frame when run through Apple Compressor. If it doesn’t, then bully for you! You can stop reading.

But if you’re afflicted with this issue, there’s an easy workaround. If you add any video track, even a text track, to the quicktime file using QT7 Pro, it will pass through Compressor like a dream (with a slight gamma shift). I’ve made a quicktime text track of a tiny period character in the upper-left corner that I paste into Quicktime files for this purpose.

To use:

  1. Download this zip file (local link, dropbox link) and uncompress it.
  2. Open the quicktime “dot-text-track-invisible.mov” in QuickTime 7 Pro (assuming you can still run QuickTime 7 Pro — at some point a macOS update is going to break that ancient piece of very usable software, and I don’t think QuickTime X lets you add tracks to a QuickTime file).
  3. copy to clipboard
  4. open your dnxhd quicktime in QuickTime 7 Pro
  5. From the menubar, choose “Edit > Add to Selection & Scale”
  6. Save the dnxhd quicktime, which now has a new but invisible text track composited atop its video track

The resulting quicktime file should now be fully usable in Compressor. Having more than a single track in the file somehow knocks the compression job into a different code path or portion of the compositing engine that avoids the problem, or maybe it just flips a magic bit somewhere, I know nothing beyond speculation –but it works.

In case you’re reading this and thinking,”Why would I use Apple Compressor for this when I have Adobe Media Encoder which is not afflicted by this DnxHD issue?”, the answer to that could be any of the following:

  • A properly-configured Apple Compressor on a multiple-core machine (or network of machines) still beats Adobe Media Encoder on the same machine at making H264 in both speed and quality, at least last time I tested it.
  • You may already have created some nice presets for making files with certain watermarks in Apple Compressor and want to use them.
  • One click to compress and author/burn a simple DVD or Blu-ray disc or image. Nice.
  • You don’t have the Adobe Creative Suite on hand and are intimidated by ffmpeg

As an aside — ffmpeg can do watermarking, scaling, timecode burns, and a billion other things (as well as encode to Prores on Windows machines), and I highly recommend it. But it’s got a steep learning curve once you get into complicated tasks.

 

Me on Mastodon.

Sign of the times – firmware updates for cars and pacemakers

Wednesday, September 13th, 2017
Retrieving the odometer reading using a car iphone app.
Retrieving the odometer reading using a car iphone app.
The firmware update message that appeared on my friend's car's display last night as we headed out on an errand.
The firmware update message that appeared on my friend’s car’s display last night as we headed out on an errand.

Yesterday I ran into 3 signs that we live in the future.

  1. I had to report my wife’s car’s odometer reading to our car insurance provider to help estimate mileage for policy renewal. She was busy and I had no physical access to the car –but I realized during the call that I could launch her car’s dedicated iPhone app, log in to her car, and fetch the desired information. Crisis averted.
  2. A friend’s car offered up the option to upgrade its firmware while we were off running an errand.
  3. And this is kind of a cheat since it’s not a personal observation – but last night I came across this blog post by JWZ about a “Firmware Update to Address Cybersecurity Vulnerabilities Identified in Abbott’s (formerly St. Jude Medical’s) Implantable Cardiac Pacemakers”. That would surely make me nervous if my life depended on one of those devices, both the prospect of having a mechatronic body part hacked and also the prospect of surviving the firmware update.

Getting OS X Spotlight to index NTFS drives

Wednesday, June 7th, 2017

I had a need to select all quicktime files on an NTFS drive to open in a particular application (MetaCheater v2.0, if you must know), and the built-in file search feature in OS X called “Spotlight” seemed like the perfect solution as it was built into the file selection dialog that the application presented. But it turns out that OS X did not index the NTFS drive I’d connected, even though I’d mounted it using Paragon’s NTFS for Mac pane in System Preferences. But a google search turned up this tip from ‘Dracos’ on the macrumors forum.

Go to Applications – Utilities – Terminal

Type in:


/usr/bin/mdutil -i on /Volumes/NTFS_DRIVE
(Enter)
/usr/bin/mdimport /Volumes/NTFS_DRIVE
(Enter)

That worked!

Just filing this away in case I ever need it again.

Converting mkv, mpeg-2, anything from disc, to DnxHD or Prores using ffmpeg

Friday, March 25th, 2016

Let’s suppose, for the sake of argument, that you’ve been asked by a director to cut his directing reel. He hands you a DVD and a Blu-ray and asks you to grab clips from them.

There are three problems here. The first is that you really should have been handed Prores or something editable from the get-go. But let’s assume you can’t get the media any other way (because often, you just can’t), and let’s assume that this is all considered fair-use (though current law makes that a bit muddy). We’re left with two technical problems:

  1. Rip: Getting the media off the disc and into a file, which may require circumventing copy protection mechanisms and encryption.
  2. Transcode: Converting the resulting file into something editable. Avid Media Composer won’t cut a .mkv file, and works with a limited range of video codecs. FCPX and Adobe Premiere Pro, despite their focus on working efficiently with many codecs, will still have trouble with some file formats. So this step is required for them too.

Rip

Regarding getting the clips off the disc, there are several applications out there for DVD, and only one that I know of for Blu-ray. From DVD, many people use a pretty fancy open-source tool that lets you pull selected ranges of video, and saves to mp4 (the software is designed to make files that play back on media devices, not for editorial purposes). If I’ve ever had to do this hypothetical task, I’ve avoided using this tool because I want the video to transfer from disc to file with no transcoding or lossy compression in between. So I’d have used this one tool which works for both DVD and Blu-ray. The result is a .mkv file which contains all the video and audio from the disc in its original codec.

So you’ve ripped the video from the disc and you’ve got an mkv file or files. You can play these in VLC Player to have a look at their contents to be sure you’ve got what you need. But none of the NLEs with which you edit can work with the files directly. Here’s how to convert them to dnxhd in order to be able to bring them into Media Composer. We’re going to use an awesome command-line tool called ffmpeg.

Install ffmpeg

At this point, I’m going to be giving instructions that relate to computers running Apple’s OS X. I’m sure it’s possible to do the next steps on a machine running Windows or Linux, but I’m not sure of the exact process. And we’re going to be using the Terminal, which you can find in /Applications/Utilities. You might as well open a Terminal window now.

There are a number of precompiled versions of ffmpeg floating around for download, but the only way to be sure you’re getting one with support for all the necessary formats built-in is to build it from source on your machine. To compile it, you’ll need Apple’s free development tool, Xcode, or at least the Xcode command line utilities. Download and install that from Apple. Then you’ll need to install ffmpeg and its dependencies, a task that is made pretty easy by a packaging system called ‘homebrew‘. If you’ve never set a password for your user account on OS X, you’ll need to do that first (the unix ‘passwd’ command, entered at the Terminal prompt, is a fast way to do this). It’s very easy to install homebrew, you just copy and paste one long command from homebrew’s web site into your Terminal window.

Once you’ve got Xcode and homebrew installed, you’re ready to install ffmpeg. I do so by pasting the following into the terminal window:

brew install ffmpeg --with-fdk-aac --with-ffplay --with-fontconfig --with-freetype --with-frei0r --with-libass --with-libbluray --with-libcaca --with-libquvi --with-libsoxr --with-libvidstab --with-libvo-aacenc --with-libvorbis --with-libvpx --with-opencore-amr --with-openjpeg --with-openssl --with-opus --with-rtmpdump --with-schroedinger --with-speex --with-theora --with-tools --with-x265

That should all be one long line. Hit return, and brew will start downloading tool after tool, occasionally prompting you, and eventually installing ffmpeg. This may take a while, depending on the speed of your computer and the speed of your net connection. A half-hour wouldn’t be out of the ordinary. But when it’s done you’ll have an amazing multimedia swiss army knife at your disposal. I use ffmpeg for nearly all of my transcodes, even to watermark and transcode for upload to video distribution platforms.

Transcode your files with ffmpeg

Now that ffmpeg is installed, you can transcode your mkv file into dnxhd with one line of commands entered at the Terminal prompt. Here it is:

ffmpeg -i inputfilename.mkv -vf "scale=1920:1080" -vcodec dnxhd -r 23.976 -b:v 36M -c:a pcm_s16le -ar 48000 -ac 2 outputfilename.mov

For command-line newbies, I should point out that spaces are used as a special character to separate commands and their arguments. So if there are any spaces in your filename, or in any directory along the way to your file, they need to be typed in a special way (“escaped” with a  backslash in front of it, like “\ “) so that the Unix shell knows to interpret them as a space and not as a separator. There are other special characters too. The safest way to type the filename is to let the Terminal do it for you — just type “ffmpeg -i ” (don’t type the quotes, but do type the space after the i) and then drag your input file right into the Terminal window. It’ll type the whole path to the file into the Terminal, with all spaces and special characters properly escaped. So for this command, I do:

ffmpeg -i drag_in_mkv_file -vf "scale=1920:1080" -vcodec dnxhd -r 23.976 -b:v 36M -c:a pcm_s16le -ar 48000 -ac 2 drag_in_mkv_file_and_then_edit_its_name_to_end_in_mov

Hit enter, and ffmpeg do its thing. It’ll take the video from that mkv file, uprez it to 1080p if necessary, and output it as a DNxHD36 23.98 quicktime file with stereo 16-bit 48Khz audio. If that matches your Media Composer project’s settings, the video will fast-import. Unfortunately I haven’t yet nailed down the exact audio settings, so that audio will just do a normal and not fast-import (which means some conversion is happening during the import).

If the file you’re converting isn’t 23.98p, change that part of the command line to the appropriate frame rate. You’ll likely also need to change the bitrate. There’s a list of bitrates to use for different frame rates of dnxhd here (scroll down to the dnxhd section).

You can also convert your mkv or other files to prores using ffmpeg. For our blu-ray mkv file, a line like this should work:

ffmpeg -i inputfilename.mkv -vcodec prores -profile 2 -pix_fmt yuv422p10le -acodec pcm_s16le outputfilename.mov

That’ll make a Prores422 file with 16-bit audio. You can also do other types of Prores by changing the number after “-profile” to 0 for proxy, 1 for LT, or 3 for 422HQ.

In summary

Yes you can pull clips from DVD or Blu-ray and convert them into editable media without an intermediate transcode to mp4 – by losslessly ripping to mkv and then converting to Avid DNxHD or Prores using ffmpeg.

And if you get heavier into the command-line, you’ll find that it makes complicated scripting and batch processing pretty easy to do. So having a good command-line conversion tool becomes very powerful.

Simple workaround for Media Composer’s ADM_COREAUDIO_UNDERRUN error

Wednesday, December 9th, 2015
Every once in a while I’d sit down at my editing system, would open a clip in the source window, would hit play, and would be presented instantly with a little popup modal dialogue window displaying the error:
• Exception: ADM_COREAUDIO_UNDERRUN, Detected:2147483648
I’d hit the “OK” button, hit play again in the source window, and the error would happen again. And again. At this point Media Composer just wouldn’t let me play anything as it kept popping up audio errors each time I hit play.

It seemed likely to me that the audio output engine had gotten itself into an unstable state.

I like trying to guess what’s going on under the hood, as sometimes it helps point the way to a solution. Looking at the error, “ADM” probably means something like “Avid Device Manager”, “COREAUDIO” is an OS X framework for audio, and “UNDERRUN” might be the type of error coming from the Coreaudio framework, complaining that Media Composer isn’t feeding it enough audio sample data. A “buffer underrun”.

It’s as if the operating system is presenting the application with a spoon (AKA the “buffer”), the size of which is directly related to the sampling rate and bit depth of the audio output setting, and the application needs to exactly fill that spoon when presented in order for any sound to be output. Media Composer for whatever reason isn’t completely filling the spoon. Maybe it’s forgotten the size of the spoon, or needs its eyes checked.

Restarting the application would probably fix the problem, and barring that, restarting the computer might fix all. A bit of shallow net searching 1 comes up with those suggested solutions. I’ve found another way that requires no relaunches or reboots:

  1. Open /Applications/Utilities/Audio MIDI Setup.app
  2. If necessary, hit command-1 to open the “Audio Devices” window.
  3. Click to select the audio output device you’re using for Media Composer, in my case at this station it’s “Built-in Line Output”.
  4. In the pane at right, switch the output format from its current channel number and bit depth setting to one of the other options in that menu. Then change it back. In my case, I switch it from “2ch-24bit Integer” to “2ch-16bit Integer” and then back to “2ch-24bit Integer” (I want to leave it at 24bit, as that’s the bit depth of my project’s audio and I’d like to avoid having either Media Composer or Coreaudio have to waste effort doing any conversions).
  5. That’s it, you’re done. Now go back to Media Composer, hit play, and bask in the glory of functional audio output.
2015-12-09 at 10.34If I were to hazard a guess as to what’s going on inside Media Composer that makes this work it is that step 4, changing the audio output device’s bit depth, pulls the rug out from under Media Composer a bit. OS X likely informs the application that the bit depth for the Coreaudio output device has changed. Media Composer probably then closes its own internal output device object (managed by the aforementioned “ADM”) that had been connected to Coreaudio and then creates a new internal output device with the new settings and connects it to Coreaudio.

So it’s kind of a reset of the application’s audio output system, and now it’s got a nice new empty set of spoons coming in, and once again knows their exact size. –Apologies to all at Avid if this isn’t correct, I’m just guessing.

A bit of google searching comes up with a similar error message afflicting those running Media Composer on Windows. I can’t offer up a solution for those unfortunate souls 2 , but I’d bet there’s a similar bit of OS-level audio setting rejiggering that will have a similar effect.

  1. Admittedly not a very deep search. It’s possible this same solution is out there in one of the forums. ↩︎
  2. Unfortunate since you’re afflicted with this error — not because you’re running that other OS that’s also not NetBSD. ↩︎

Fantastic new Star Wars trailer

Thursday, May 21st, 2015

Kudos to the VFX artist behind this new Star Wars Trailer.

An ffmpeg recipe to concatenate multiple quicktime files and convert to h264

Tuesday, April 29th, 2014

If you ever find yourself with a set of same-as-source quicktime files that have been exported from Media Composer, one per reel of a film…

–and you need to output a single h264 file of the film to upload to Dax or Pix or another of the various secure media platforms…

–and you know your way around the Unix command line and have ffmpeg installed on your machine…

–then you’re in luck. Here’s how to use these tools do what you need.

First, construct a text file that lists the files you mean to compress. My “mylist.txt” file contained the following lines:

file 'REEL 1V7 042914.mov'
file 'REEL 2V11 042914.mov'
file 'REEL 3V7 042914.mov'
file 'REEL 4V8 042914.mov'
file 'REEL 5V6 042914.mov'
file 'REEL 6V7 042914 fix.mov'

Put the mylist.txt file in the same directory as the source files, and also use the ‘cd’ command to switch your shell session into that directory so that long paths won’t be needed. Then, run this ffmpeg command (I wrote it as one line, but broke it up here with the ‘\’ character at the end of each line so that it’d be readable):

time ffmpeg -f concat -i mylist.txt -framerate 23.976 -s hd720 -c:v libx264 \
-preset slower -profile:v baseline -g 24 -deblock 0:0 -c:a libfaac -b:a 96k \
-movflags rtphint+faststart -vb 2500k -pix_fmt yuv420p -f mp4 -threads 80 outputfilename.mp4

45 minutes after I did this, I had a single mp4 file containing all the reels in order, as h264. I checked all the transition points between reels and there were no glitches.

This saved me from having to stay at work an extra 4 hours that night, waiting for Media Composer to export a full-length same-as-source movie and then converting it to h264.

I know what you’re thinking, “Why not just export to h264 from within Media Composer?”. Two reasons: Going to h264 directly from within Media Composer is slower than exporting same-as-source and feeding it into ffmpeg, and Media Composer’s h264 output doesn’t support as many options for hinting for streaming as were required for that upload.

What’s likely coming in the next GoPro (Hero 4) – 4k 30fps, 2k 120fps!

Tuesday, March 18th, 2014

If the specs for the next version of the system-on-a-chip system used in the GoPro Hero 3 are any indication, the next GoPro is going to have some ridiculous specs. 4k-resolution at 30 frames-per-second, 120fps 2k, 240fps 720p, better dynamic range, quality, etc etc etc. The Ambarella A9 sounds like a monster.

I’m looking forward to the cameras the next couple of years will bring. Now if I could just focus on writing something worth shooting…

How to rank the latests slate of Amazon pilots?

Friday, February 28th, 2014
2/27 stars, votes, and imdb rating of Amazon's 2014 pilots
2/27 stars, votes, and imdb rating of Amazon’s 2014 pilots

It’s nearing the end of the first month of Amazon’s 2014 pilot season. Amazon, being a company of the digital age, posts all the pilot episodes of their new slate of shows online (here’s a link to view all of them for free) and then tracks how well they’re received by the public. Presumably, the shows that get the most interest and the highest ratings are the ones that Amazon will decide to run for a full season, though I wonder how much or whether the expected production budget of each show figures into the calculation. I worked as assistant editor on The Rebels, so I’m likely a bit more curious than the average person as to how the various pilots are doing. I’ve checked in on the pilots roughly once a week to record their metrics.

There are a few ratings for each pilot that are publicly available. These include the star rating (out of five, rounded to the nearest half-star), number of votes, and imdb rating, listed at the top of each pilot’s page. Further down the page, in the ‘Customer Reviews’ section, is a more precise star rating rounded to the nearest decimal point (i.e. 4.7 rather than 4&1/2).

At right is a table of the current standings of Amazon’s non-kid-oriented pilots as of yesterday, and below are a few graphs showing the trend in those metrics over the past few weeks:

Amazon pilots 2014 trend graphs
Amazon pilots 2014 trend graphs

It looks to me as though all the pilots have been well received by the very non-random-sampling of Amazon members who choose to both watch and review them, with the minimum star rating hitting 4 out of 5. Only Amazon knows how many viewers each episode has accrued. Bosch appears to be the runaway hit, though I can’t say for certain that this means it’ll be picked up as there are a lot of factors to which I’m not privy –such as the budget for the show and its viewer breakdown. It interests me that The After and Transparent started strong but have seen an fall in their star (and for Transparent, also its imdb) ratings over time. Until I get around to watching all the pilots I keep in mind that it’s possible that this just means the people who are most likely to enjoy those shows are the ones who were first in line to watch them, and the downward trend has nothing to do with how much I will like them. The Rebels seems to have had a pretty consistent reception, with a center of the pack star rating and a steadily rising IMDB rating. The number of votes for The Rebels and Mozart In the Jungle have been almost identical at every moment I’ve sampled, so that’s why they obscure each other in the “number of votes” chart below.