Recordings
From ICELabWiki
This page is a list of all the locations of the recordings we have made of AG (and other) events, as well as some instructions about how to effectively transcode/format-shift certain things.
Contents |
[edit] AGVCR to MPEG2 trascoding
This portion of the wiki has a how-to for changing AGVCR recordings into MPEG2 format.
http://wiki.archer.rit.edu/wiki/Access_Grid_Recording_Post-Production
[edit] DVD & Other High-Quality
[edit] Overview
Our goal is to have 3 levels of quality for archival, sourced from high quality (ie not from AG) recordings:
- DV, 30 Mbps (or other high-bitrate, intraframe-only-compression codec)
- MPEG-2, 5-10 Mbps
- MPEG-4, ~1.5 Mbps
[edit] Ripping DVDs
The current method I'm experimenting with for crudely ripping DVDs is:
- Copy vobs to hard drive
- cat *.vob | mencoder - (-ffourcc mpeg) -of mpeg -ovc copy -oac copy -o out.mpg
- Explanation of these opaque looking options:
- ffourcc forces the file's fourcc (codec identifier, basically) to be 'mpeg'. if you don't do this, there is a broken/absent fourcc and some players refuse to play such files.
- This might not be necessary, since I don't think mpegs (as in the mpeg container format) need a fourcc. I will test this.
- It's not necessary - this (without ffourcc) works and makes a proper MPEG file that can be played, and sought within, by VLC and mplayer.
- of forces it to output in mpeg container format
- oac copy and ovc copy say "no transcoding"
- ffourcc forces the file's fourcc (codec identifier, basically) to be 'mpeg'. if you don't do this, there is a broken/absent fourcc and some players refuse to play such files.
Basically this just shoves all the vobs into one MPEG video. Note that this loses chapter info/breaks, probably subs, and I don't know what kind of nonsense would result if the vobs had multiple audio tracks.
[edit] To MPEG-4
After that, to get to MPEG-4, using VirtualDub might be the most intuitive tool to use, and it gives easier-to-see-and-use control. On Linux machines, doing all those command line options in mencoder is a bit of a hassle, hard to learn and easy to screw up, but it seems to work pretty well. It's probably easier to automate too.
I'm thinking that we should avoid going from MPEG-2 to MPEG-4 if possible, and instead go from DV to MPEG-4, since you obviously lose quality each time you transcode (this would skip any compression artifacts piled on from MPEG-2). Beyond that, it might be easier to go from DV (hopefully in an avi file) to MPEG-2 (using a modified VirtualDub: [1]) if the DVD ripping method doesn't work or if there is no DVD. Of course this is all moot if we don't have access to the DV tape, or if it doesn't exist.
The method I am currently testing with mencoder:
- 1 Pass:
mencoder input.mpg -oac mp3lame -lameopts preset=standard -ovc xvid -xvidencopts bitrate=1500 -o out.avi
- On mushrik, this took about 80 minutes for a 66 minute mpg (~25fps)
- 2 Pass:
mencoder input.mpg -oac mp3lame -lameopts preset=standard -ovc xvid -xvidencopts pass=1 -o /dev/null mencoder input.mpg -oac mp3lame -lameopts preset=standard -ovc xvid -xvidencopts pass=2:bitrate=1500 -o out.avi
- On mushrik, the first step took ~40 minutes for 66 minute input (~45fps) and again 80 minutes for the second step (25-26fps).
- Don't worry about the /dev/null thing, this first pass generates a divx2pass.log which the 2nd pass uses.
I'm also experimenting with MPEG-4 inside a proper mp4 container instead of an avi (since streaming avi is iffy in vlc - seeking works on some platforms but not others). The method is:
- 1 Pass:
mencoder input -of lavf -lavfopts format=mp4:i_certify_that_my_video_stream_does_not_use_b_frames \ -oac mp3lame -lameopts preset=standard -ovc xvid -xvidencopts bitrate=1500:max_bframes=0 -o output.m4v
- 2 Pass:
mencoder input -of lavf -lavfopts format=mp4:i_certify_that_my_video_stream_does_not_use_b_frames \ -oac mp3lame -lameopts preset=standard -ovc xvid -xvidencopts bitrate=1500:max_bframes=0:pass=1 -o /dev/null mencoder input -of lavf -lavfopts format=mp4:i_certify_that_my_video_stream_does_not_use_b_frames \ -oac mp3lame -lameopts preset=standard -ovc xvid -xvidencopts bitrate=1500:max_bframes=0:pass=2 -o output.m4v
The b-frame nonsense is necessary because lavf (what you need to use in mencoder to make non-avi or mpg container files) can't handle b-frames for some reason.
[edit] Locations
All prominent AGVCR (and other) recordings are now on LMC. In the future they will probably be moved to a dedicated disk/web services server. There are a few errant AGVCR recordings, all on delilah (the folder "AGVCR Recordings" on the desktop - this is mirrored on the laptop), which are all from before spring quarter and a number of which are encrypted.
[edit] List of recordings we have archived
[edit] AGVCR
- 2/15/07 - RC Seminar, Dave Matthews
- 4/5/07 - RC Seminar, Carl Davis
- 4/19/07 - RC Seminar, Rob Stevens
- 5/17/07 - RC Seminar, Dina Newman
- 5/23/07 - RC Seminar, Paul Avery, Open Science Grid
- 5/23/07 - Demo for Dean's Council
- 5/24/07 - Virtual Theatre Demo
- 6/1/07 - Motion and Sound
- 6/8/07 - U. Manchester Seminar
- 6/14/07 - RC Biweekly Meeting (with Teragrid presentation by Paul Mezzanini)
- 7/31/07 - CIS/MCSL meeting with Mark Fairchild & President Destler
- 8/13/07 - Multicast Networking Overview
[edit] Recorded Streams
- 5/25/07 - 2007 RIT Commencement
- recorded MPEG-2 of main ceremony starting from after the national anthem to end, experimental transcoded theora version from aforementioned start point to a bit after Clinton's speech, and recorded MPEG-2 of the full NTID commencement
[edit] DVD & Other
- MPEG-2 & MPEG-4 (XviD 1-pass and 2-pass) videos of Paul Avery's Open Science Grid lecture, sourced from the ETC DVD
- XviD 1-pass and 2-pass conversions of Convergence (Research Channel educational video)

