Slashdot Mirror
Slashdot Asks: How To Best Record Remote Video Interviews?
You've probably noticed that Slashdot's been running some video lately. There are a lot of interesting people and projects in the world we'd like to present in video form, but some of them are too far away for the corporate overlords to sponsor travel to shoot footage in person. (Another reason my dream of parachuting to McMurdo Station will probably never manifest.) We've been playing around with several things on both the software and hardware side, but in truth, all of them have some flaws — whether it's flaky sound (my experience with the otherwise pleasing RecordMyDesktop on Linux), sometimes garbled picture (Skype, even on seemingly fast network connections), or video quality in general. (Google Hangouts hasn't looked as good as Skype, for instance. And of the webcams built into any of the laptops we've tried, only Apple's were much worth looking at. Logitech's HD webcams seem to be a decent bargain for their quality.) We've got a motley bunch of Linux, OS X, and Windows systems, and can only control what's on our side of the connection: interviewees may have anything from a low-end laptop with a built-in webcam to elaborate conferencing tools — which means the more universal the tools, the better. (There may not be any free, open source, high-quality, cross-platform video conferencing tools with built-in capture and a great UI, but the closer we can get, the better.) With all that in mind, what tools and workflow would you suggest for capturing internet conversations (with video and sound), and why? Approaches that minimize annoyance to the person on the other end of the connection (like the annoyance of signing up for an obscure conferencing system) are especially valuable. We'd like to hear both sides, so please chime in if you've had especially good or bad experiences with capturing remote video like this.
Courtroom sketch artists
Whatever you end up doing I would appreciate it if you could include a transcript of the conversation if the video isn't particularly important. I usually end up reading these stories on break or when it is particularly slow, so transcripts make an otherwise worthless video entry useful to me.
I am curious why video is necessary for these interviews.
Sounds like you have an itch. Scratch it. You have a legitimate business need for this software, use some of that corporate money and invest in the tools you need. Then open source them so we can all enjoy.
Give me Classic Slashdot or give me death!
Forget video. Just post a transcript of the interview to Slashdot and a link or embed the youtube video. I really enjoy being able to read the TED talks because I can't stand watching most of the videos. I like reading because it's faster than listening to some one pause between concepts.
Consider holding the video conference using the lowest-common-denominator, e.g. Skype, but having the remote party also record it locally using a higher quality codec with standalone video recording app. Once the interview is over, have the remote party compress the file (Handbrake is easy to use, even for non-tech minded) and transfer it using conventional means (FTP, etc.). Though this might mean you'd need two cameras at the remote location (unless you have a way to split a video device to two different apps), it will eliminate network bottlenecks, latency, and resolution constraints.
My friend has a tech podcast he produces weekly, and he's run into the same desire to do remote interviews.
We've found that for a Windows PC, the Logitech Orbit AF was a pretty good webcam. It has motors in it so it can pan and tilt, as well as auto-focus, and the camera "ball" sits on a tall, thin stand so it's basically at eye-level when sitting on a table or desk in front of the user(s).
The stand is, unfortunately, a little on the flimsy side (basically, the ends of the plastic pole that snap into the camera and the base are just mini USB connectors - so liable to break if the camera is knocked over a few times). But if you treat it with care, it seems to be pretty effective. Logitech's accompanying software supports face tracking too - so if the user moves, the camera tries to follow their head.
As for software, my friend always used Skype - but definitely with less than stellar results. The big benefit to Skype, though, is its popularity. Most people you call and ask to do a video-conference with you either have Skype already set up, or can quickly download and install a copy. Plus, it's cross-platform compatible. But yeah, it seems like video quality varies with it, even when both parties have very fast broadband connections -- and on longer sessions, it seems to eventually lose the audio or video at some point in the call.
The best solution is to provide a software to the other end to record the conversation. That way his/her answer and behavior will be perfectly recorded and that is what you care about. Then send the file by any means (network, USPS, ...) Provided these are *slashdot* interview, I am sure the interviewee (is that actually the right word?) will know how to do that.
Maybe because text is so much more efficient to create, transmit, store, analyze, consume, etc.?
While I don't use xsplit for video conferencing, I see lots of casters use it to stream content (from web cams, desktop, to direct x games). I tried it out and loved the result. In the end, I did not like the sound of my recorded voice so I stopped casting.
I see the pro version supports skype. It would be worth a try or awesome if someone that had experience with that feature could comment on it.
http://www.xsplit.com/
Best I've seen so far has been Apple's iChat. It lets you record video chats natively and without hassle, with both parties having to agree.
Works with AIM, Jabber/Google Talk, and (in the Messages beta) iMessage.
Obviously the main flaw is that it doesn't work if the person on the other end doesn't have a Mac. If they do, though, I don't see any reason not to use it. There are even scripts you can use to set it to automatically ask to record each call when it starts.
Dan Aris
Fun. Free. Online. RPG. BattleMaster.
"We'd like to hear both sides" haaahaahahaa
I'm tired.
Karnal
Body language
Tone of voice
Getting an idea of the person being inteviewed
Hearing a human voice in this often written world
These are reasons why a video and transcripts are different, not why either is better. Any one of those can skew the usefulness or appeal of a message being presented by video. Of course, sometimes the message itself is not the primary point, and the ability to make judgements about the messenger is paramount. IMO, video should only be used when the credibility of the messenger is significantly important. Otherwise, it is a waste of resources.
Wowza is low cost, supports AWS hosting and has DVR functionality built-in, you just point the client at the DVR App to start recording broadcasts. It will save this broadcast as an FLV file on the server, one for each broadcast. You can extend this functionality to whatever edge case you may have by using Java POJO coding as well as transcode using FFMPEG. Another nice plus is that it can support playback at various bitrates so you can offer adaptive playback to support lower bandwidth clients as well as those with fast connections. On the whole, it's very straight forward to get up and running with Wowza.
The draw back is that it uses RTMP for incoming stream recording and since there is no HTML-based Device support (yet), use of a webcam or external camera requires you use Flash based technology for the Broadcast Client. The good news is that you can develop a Flash based client for all platforms, aside from Mobile Browser. Android, iOS and WebOS can be delivered as Native Apps written with Flex with a captive Adobe Air Binary inside, Air also allows you to create Desktop Apps for Windows and Mac, though Linux support has been tossed due to lack of interest from Flash hating Linux users (however Linux support can be had by using older SDKs if it is an absolute must have). Wowza supports re-streaming, so the videos could be streamed as mp4 instead of flv and via other protocols (such as HLS) so other non-Flash based players (such as an HTML5 video tag in a mobile browser) could play back the recorded video.
The major cost here is in the development of a tailor-made broadcast and playback client to the various platforms (a good AS3 developer could write a single core codebase that has a web, and native app UI to minimize cost). The Wowza license is $50/month on AWS and similarly priced to host on your own hardware. Unless you have several hundred or more regular viewers, bandwidth costs should not be a factor.
to a text based discussion like /.?
Since when was this a thing? Sounds like something an uneducated bar star would say about finding a partner. "You have to pick any two : Rich Smart or Good Looking". Those are stupid compromises made by pessimists and defeatists.
I could name a few dozen open source projects that are all 3 but I won't. You've interacted with at least a dozen just by posting here.
I'm workign in TV news now, and have worked with many of the majors over the past few years. There is not a universal winner.
Skype does remain extremely popular because of it's ubiquity and it's fantastic ability to handle very crappy connections. It's not the best when you have plenty of bandwidth, but it's certainly the best when you don't. CNN takes live shots to air off Skype regularly.
That said, if you are willing to pay, Vidyo is doing some decent stuff. I know of at least 3 'radio shows on tv' that operate over Vidyo, with cameras at the radio studios, and the broadcast control room hundreds if not thousands of miles away. I'm not sure about linux support, but there is a free client for OSX and Windows, so any interviewee can download it and join right up using whatever webcam is available.
I'd recommend not looking for a single solution to rule them all. Instead, try and schedule about 10min at the beginning to see what you can get working - start w/ a high end, like Vidyo, if it's struggling or they can't get ti going, try facetime, still no love, step down to Skype.
That is, frankly, what the pros are doing.
If you don't want to tax the computer at all.. Split the video signal from the computer to a record deck. Say it's a VGA output. scale it to YUV and record it on a HDVcam. Then edit the video. I know this sounds like a the long way, but it will probably be the best quality.
While recording to a deck, you could also have a good camera setup that inputs to the Video conerence and splits the signal to another recorder.
Other options for recording video conferences without skype are actual conference units. like Tandberg or polycom. they usually include outputs that allow for recording. Also video conference bridges can be used for the purpose of recording the conference as well. They charge quite a bit to ensure the call, but work very well.
and there are reasons why a transcript is better:
when I have limited time and I want to read through the important parts.
When I want to have a brief scan of the topics to see if its something I want to view in depth.
If you only provide video, I have to decide whether its worth spending my time viewing it or not, and typically, that means I pass up and head for the next link. There are plenty of links vying for my attention, I don't need ones that are in less than accessible.
What methods are people using to provide these transcripts? Piping it to Google Voice? Splitting out the audio and running through Dragon Dictate? Farming out to Mechanical Turk?
I was thinking about trying to provide transcripts for a large number of videos and nothing strikes me as particularly easy or reliable. Any suggestions?
We'll be right in touch ... Anonymous :)
Seriously, pop an email to feedback at slashdot with more details, if you have any video pitches to present :)
timothy
jrnl: http://tinyurl.com/c2l8yr / foes: http://tinyurl.com/ckjno5
Try http://www.shinywhitebox.com/ishowu-hd/
10.5-10.7 support, lets you record screen, cameras and has great support.
Domestic spying is now "Benign Information Gathering"
Once you find a software solution, don't forget the hardware. I recommend shipping a Logitech or HP webcam as well as a decent audio capture device to the person to be interviewed and have them ship it back when done. Low cost, and you standardize the video and audio capture quality.
I've played around a lot with this stuff for more than a decade. Skype can be okay if the remote system is powerful, well configured, has excellent broadband, a good external webcam and good lighting. However, those necessary things will only rarely all be present, particularly internationally. If you want high quality and a high degree of future-proofness for the video assets you are going to so much effort to create, there is a counter-intuitive alternative.
There are truly amazing small consumer camcorders now available for $500-$700 that work great in low light, capture excellent 1080p, auto focus on faces, auto iris, auto WB, etc quite well. I shoot with high-end pro gear and these little consumer cams deliver an unbelievable picture for the price. Ship one of these to your remote location with a small AV clamp that will mount it to the top of the monitor next to the webcam. Do your live interview via Skype through the webcam but have your remote location turn on the HD camcorder after the Skype session begins. The camera will beep when it starts recording and you can use this beep to later sync the Skype recording with the HD camera recording. After the interview, the remote location can plug the camera in as a USB device and Dropbox the recorded file over to you in non-real-time (AVCHD peaks at ~24mbps but is often much smaller). Or if it's not as time sensitive, you can get the file off the SD card when the camera comes back.
This also gives you the advantage of providing a handheld camera to your remote location. Even rank novices can shoot 'B-Roll' type footage of remarkable quality. The handheld image stabilization on these cams is impressive. If you give your remote amateur "crew" a simple shot list and ask them to first watch a ten minute YouTube tutorial on basic camera handling and shot composition, the results can even rival semi-pros. This way you'll have something to intercut with the talking heads to further elevate your production value.
Yes, this implies that you are actually editing your final product. Basic editing will again double your production value over raw webcam recordings. It doesn't have to take too much extra time, particularly if your remote camcorder has it's time-of-day clock set roughly right. Your handy intern can be making a shot list of good/bad clips and restarts on your end during the interview and reference this via TOD + clip offset time during subsequent editing. This saves a lot of hunting around inside the clips. With practice you should be able to do a basic edit with canned intro/outro, standard title overlays, B-roll cutaways and some still graphics (logos, product shots) in about 3X the total running time of your output. Note: that's just active work time, not unmanned background clip downloading or final output encoding, which you can batch up and leave running unattended.
Finally, as we say in TV land, audio is more than half the picture. Sending a basic wired clip-on microphone will do wonders for your production value by reducing machine noise and ambient room echo. If your remote location is in North America you can send them the Radio Shack part number for a decent clip-on mic that they can pick up themselves for $25. Regardless, if sound/video from your end is going to be seen then you should use at least use high quality mic and camera on your end.
Until Kazam gets a bit more traction and is further developed stick with FFMPEG and use a script like this:
:0.0+$WIN_XY -f webm -vcodec libvpx -threads 2 -y output.webm
#!/bin/sh
INFO=$(xwininfo -frame)
WIN_GEO=$(echo $INFO | grep -oEe 'geometry [0-9]+x[0-9]+' | grep -oEe '[0-9]+x[0-9]+')
WIN_XY=$(echo $INFO | grep -oEe 'Corners:\s+\+[0-9]+\+[0-9]+' | grep -oEe '[0-9]+\+[0-9]+' | sed -e 's/\+/,/' )
ffmpeg -f alsa -ac 2 -i hw:0,0 -f x11grab -r 15 -s $WIN_GEO -i
I seem to recall a teaching software called Blue Button, that was featured on FLOSS Weekly. Not sure if it went both ways tho.
Defining Statistics and Social Research
I highly recommend using Snapz Pro X on your end. This can record all the audio and video that shows on your computer (which must be a Mac). You would just need to setup a Skype call with your interviewee, start recording and off you go. You can also set it to record only a section of your screen (e.g., the main Skype window). I've used it to record PowerPoint lectures pretty successfully (including ambient audio).
I believe iChat can have better video quality than Skype, but it is not sufficiently cross-platform for all your interviewees. So you're probably stuck with Skype unless you want to start posting videocameras back and forth to your interviewees.
You would probably do well to make a separate recording of the interviewer using a videocamera, and splice that in to the interview.
Highly compressed audio is hugely affected by how much background noise is present in the source audio. You want to maximize the spoken voice, and minimize everything else. This optimizes the compression algorithms by giving them the best possible data upon which to base their decisions about what gets taken out of the audio. The compression ratios are often 60:1 or higher, which means that for every 1 byte of data that gets passed on to the listener, 60 bytes are taken out. Ouch.
You must get the microphone close to the speaker. Sound pressure degrades as a square of the distance from the source, so get the mic right up close. A clip-on microphone is great because the body is partially shielding the microphone from background noise, the mic can pick up sound waves coming from from the chest as well as the mouth, and it is unobtrusive.
Do not compress to a stereo codec, a mono codec will actually sound clearer. The speaking voice has a very limited dynamic range, so you don't want the compression algorithm to devote any effort to including sounds outside that range. Using a cheap microphone can actually help in this regard, as it will not even respond to frequencies at the extended ranges.
"We receive as friendly that which agrees with, we resist with dislike that which opposes us" - Faraday
Bloggingheads posts hour-long video conversations between two people and have been doing it for a few years now. I know they're now trying to use Skype when possible, but that requires high-speed on both ends, so they still use the old system a good portion of the time. My understanding is that there's some sort of hardware setup that gets shipped by post to the interviewee, and each side talks over the phone and looks at a camera. The people talking can't actually see each other, which is one downside, but the quality is generally good. http://bloggingheads.tv/