"Brian J. Murrell" wrote: > On Sun, Dec 02, 2001 at 09:24:32AM -0500, Stan Brown wrote: > Well, not really since your requirements eliminates the use of the > NVrec tools. Try mp1e. I will warn you that it has a/v sync > problems, especially when used with WMP. I used it, and looked at the a/v sync issues. The problem was that it dropped audio data at the drop of a hat. I patched my copy (1.8.1) to use real time priority, with the capture threads running at the highest priority (within the app). Another patch I made relaxed the conditions under which mp1e decided to drop frames based on the time taken to encode the current frame (I was never quite sure why the author decided to put that check in). That made almost all of the problems go away, and capture was rock solid, except that my sound card (like many others) only approximates the requested sample rate, due to physical limitations of the cheap clock crystal. I worked around that using a technique suggested by Justin Schoeman, and applied a quick and dirty stretching algorithm to the sound capture code (oss only). This is the ugliest part of what I've done, as it requires that you first run a profiling program to measure your sound card's drift from the ideal sample rate, and then plug a fudge factor into the code and recompile, to compensate. Sound quality suffers, and the patch is so ugly that I was embarassed to show it to anybody. It works for me, and for a couple other people who tried it. I used it to capture about 4 seasons of Babylon 5, with (mostly [*]) perfect a/v sync. I never had time to clean up my work, or document it, and then Justin released his nvrec package with the new capture core, which seems to be better designed, more flexible, and just all around the coolest thing since animal crackers, and I dropped my plans to write a flexible extensible, rock solid video capture library, focusing instead on cooking and lawn care. mp1e still works for me, though, and I'll be happy to give you the patched source, if you'd like. - Ori Pessach [1] One episode had the audio off by a second. I suspect that the initial a/v syncing code is sensitive to scheduling jitter, but I was never able to reproduce the problem to verify that.
