Tweet
Guys,
Rob Tywlak (my co-mod at the DMN forum) was first to notice this problem and brought it to the attention of the group. Since then he and I have spent two days, and two very long nights, tracking it down.
The problem manifests as I-frames in an encoded MPEG that have a pronounced increase in pixelization. This causes the resulting video to visually "pulse" every 1/2 second. These pulses don't occur throughout the whole video but have to be triggered by another factor. More on this later. There also appears to be a bitrate overrun.
At any rate here's the rundown;
1. I could only get the pulsing to show up at very low bitrates, which is why I didn't see it initially, but Rob can get them to show up at medium to high bitrates. Still trying to work through that one.
2. It may be difficult to see the pulsations in live video with some motion, but it shows up very well when you use a bitmap as a background....especially something with lots of detail (trees etc.) and a stationary text title.
3. In my experiments the pulsing only occurs after a transition that uses pixel rendering; basically those that scale, rotate, fade, color or do something other than just break up the frames and move their contents around linearly in the horzontal or vertical planes.
Examples of those transitions that initiate the problem are crossfade, burn, barn doors, flying box, rotations etc.
Examples of those transitions that do not initiate the problem are checkerboard, slides, builds, wipes etc.
4. Overlays that fade in/out using keyframed transparancy do not initiate the pulsations. This includes titles. This tends to indicate that fading isn't a problem in and of itself but that it's somehow related to how transitions are handled combined with an encoder issue.
5. In another wrinkle the pulsations only last 10-11 seconds after the end of a transition on my system but last for the duration of the project in Robs. Go figure.
6. The pulsations and bitrate overrun can be stopped by replacing ONE FILE in C:/Program Files/Common Files/Ulead Systems/Mpeg/ with its counterpart from the MSPro 7.01 patch. The culprit appears to be the file named MPGVOUT.004
Unfortunately there is a cost to replacing this file with the 7.01 version: you lose 2-pass MPEG encoding. The encoder goes through the motions of doing a 2-pass encode, but doesn't actually perform one.
A full report has been sent to Ulead and the beta team along with project and data files to reproduce the problem.
Ulead replied last night that our efforts have pointed them to what they believe is the problem and that they are working hard on correcting it ASAP.
Just call me;
DocM, "the sleepless" :P
Rob Tywlak (my co-mod at the DMN forum) was first to notice this problem and brought it to the attention of the group. Since then he and I have spent two days, and two very long nights, tracking it down.
The problem manifests as I-frames in an encoded MPEG that have a pronounced increase in pixelization. This causes the resulting video to visually "pulse" every 1/2 second. These pulses don't occur throughout the whole video but have to be triggered by another factor. More on this later. There also appears to be a bitrate overrun.
At any rate here's the rundown;
1. I could only get the pulsing to show up at very low bitrates, which is why I didn't see it initially, but Rob can get them to show up at medium to high bitrates. Still trying to work through that one.
2. It may be difficult to see the pulsations in live video with some motion, but it shows up very well when you use a bitmap as a background....especially something with lots of detail (trees etc.) and a stationary text title.
3. In my experiments the pulsing only occurs after a transition that uses pixel rendering; basically those that scale, rotate, fade, color or do something other than just break up the frames and move their contents around linearly in the horzontal or vertical planes.
Examples of those transitions that initiate the problem are crossfade, burn, barn doors, flying box, rotations etc.
Examples of those transitions that do not initiate the problem are checkerboard, slides, builds, wipes etc.
4. Overlays that fade in/out using keyframed transparancy do not initiate the pulsations. This includes titles. This tends to indicate that fading isn't a problem in and of itself but that it's somehow related to how transitions are handled combined with an encoder issue.
5. In another wrinkle the pulsations only last 10-11 seconds after the end of a transition on my system but last for the duration of the project in Robs. Go figure.
6. The pulsations and bitrate overrun can be stopped by replacing ONE FILE in C:/Program Files/Common Files/Ulead Systems/Mpeg/ with its counterpart from the MSPro 7.01 patch. The culprit appears to be the file named MPGVOUT.004
Unfortunately there is a cost to replacing this file with the 7.01 version: you lose 2-pass MPEG encoding. The encoder goes through the motions of doing a 2-pass encode, but doesn't actually perform one.
A full report has been sent to Ulead and the beta team along with project and data files to reproduce the problem.
Ulead replied last night that our efforts have pointed them to what they believe is the problem and that they are working hard on correcting it ASAP.
Just call me;
DocM, "the sleepless" :P
Comment