Scene Detection Abnormalties

jbar wrote on 8/18/2003, 9:51 PM

I see that Videocanuck is also asking about Scene Detection Problems,
but I'm posting this separately in case there are different issues.

Capturing from a Digital8 I'ved noticed that Scene Detection doesn't work properly, or so it seems in 2 different instances:

1. 2 scenes separated by a blank spot on tape causes all 3 pieces to be grouped into 1 avi file. Is this the known "striping" issue ?

2. 2 scenes where the 2nd scene has an older timecode/date than the 1st

These were from 2 different tapes. All other detection was accurate.

What really bugs me about this is that I used Scenalyzer (free version) to find out which clips were correct ! I paid $$$ for Vegas and it can't even do simple splits based on time or lack of ? I don't get it. I hope its a bug (that will be fixed).

Vegas 4.0c (did 4.0d improve capture any ?), Win 2k Pro

Comments

Former user wrote on 8/19/2003, 8:44 AM

In reference to the blank between scenes. What do you have your MINIMUM scene length set for? It could be the blank was below the minimum so it just kept it as one scene.

2: Just curious, how did you get a scene with an older time/date stamp AFTER a later scene?

jetdv wrote on 8/19/2003, 9:01 AM

2: Just curious, how did you get a scene with an older time/date stamp AFTER a later scene?

That would be EASY - reuse the first portion of the tape. The front portion will have a NEWER date than the last portion!

Former user wrote on 8/19/2003, 9:09 AM

Kind of risky if you actually want the earlier footage. ;)

Dave T2

Randy Brown wrote on 8/19/2003, 9:47 AM

I find it will split my "scenes" into different clips. Not a big deal since they play flawlessly on the timeline and if I need to (to open the audio in Sound Forge), I can unsplit them with Excalibur; a minor inconvenience, just curious as to why.
Randy

jbar wrote on 8/29/2003, 10:48 PM

I finally got some time to investigate this further.

I've been mixing up some terms. "Timecode" seems to be specifically referring to an hh:mm::ss;ff value recorded onto the tape. I had been under the assumption that the date/time on the tape was the timecode. So in my previous post, all references to timecode was not to the "real" timecode. Scenalyzer can parse on date/time, which is what I thought VV was doing. But VV seems to ignore the date/time, when it comes to detecting scenes.

In problem #1, the blank spot caused the timecode to start over at 00:00:00;00, which caused VV to skip over that scene change, even though the time/date changed.

For problem #2, I had previously recorded something over the entire tape. Then re-used the tape and at one point, part of the original recording was left. When I referred to an older timecode, it was actually an older time/date. The timecode was still linear.

So I'm left with several questions:

1. If VV only uses the timecode to detect scene changes (true?), what indicates the scene change ? Hidden marker ? Maybe a flag when recording starts ? This would explain why in #2, the 4 seconds of original recording was left appended to another clip.

2. From problem #1, now that I have a tape with the timecode split (starts over half way through), is that permanent ? Since in problem #2, the timecode seems to have been left intact even after re-recording. Or if the timecode was overwritten, what's to keep the same timecode from being used twice if the timing between the two recordings is just slightly off ?