PDA

View Full Version : Output Rate Formula

Marsman2k
8th July 2004, 16:31
I see a lot of people struggling with getting the desired output rate. One thing that I don't think many people have realized yet is that the idea of adding the video rate plus the audio rate equals the output rate is only valid at 30 frames per second!
Example: for a 128k output: Video: 104 / Audo: 24 = 128k <- but ONLY at 30 fps!

For other fps rates, if the desired output rate is to be maintained, the video data rate has to be increased. Lets use an example of 15 fps. Here we have exacly half the number of frames per second as with 30 fps. Since only half the frames are rendered, the output total for video is only 52k - even though we specifed 104k. This is because the video rate is calibrated for 30 fps. The picture quality will remain the same as the process always gives each frame 1/30th of the video rate, but since there are only 15 frames, the total output is one half.

In order for those 15 frames to total 104k, they need to be encoded twice as big as with 30 fps. By doubling the video rate to 208k - with each one of the 15 frame's receiving 1/30th of the data rate, the result is 104k - exactly what we wanted. Note that the picture quality is increased as each frame is given the same encoding as it would for a 30 fps @ 208k video. So each frame looks like a 208k encode, but since there are only 15 frames, the output rate is just 104k instead of 208k if there were 30 frames.

Therefore, for any encoding OTHER than 30 fps, the video + audio = output is not valid and must be adjusted. Here is a simple formula you can use to achieve the desired output:

(30/fps) x (video rate) + (audio rate) = actual output rate

Lets use our 15 fps in the formula:

(30/15) x (104) + (24)
(2) x (104) + (24)
(208) + (24)
Result: 128k @ 15 fps

So base your initial video rate on 30 fps - then use the formula to adjust the video rate to compensate for any drop in fps. Remember also that lower fps's using this method yield better video qualities as more video processing is given to each frame. I hope this help sort out some of the confusion regarding output rates at fps's other than 30.

ken52787
8th July 2004, 19:03
You sir, are incorrect. It works by kbps (kilobits per second) despite the framerate.

Out of boredom, I've decided to test your theory. All the tests use the same video clip (ripped from dvd, fairly complex) with VP61 at 480x352 with a bitrate set to 300kbps. Audio doesn't matter, but its 32kbps AAC.

So I had it encode at 30fps, 15fps, and for shits and giggles: 5fps. By your formula the results should be 332 for the 30fps, 182 for 15fps, and 82 for 5fps.

I encoded them all then got the average bitrates, which turned out to be:
30fps-324kbps
15fps-323kbps
5fps-317kbps

So as you see, the entered bitrate works by seconds, not frames.

PS: Don't take this post as hostile or rude, I'm just extremely bored and it makes me grumpy :)

Marsman2k
8th July 2004, 20:46
Originally posted by ken52787
PS: Don't take this post as hostile or rude, I'm just extremely bored and it makes me grumpy :) [/B] Hi Ken, np - my skin is relatively thick :D

I should have mentioned that my tests were done with vp62 and not 61 - apparently there is a difference in the way the encoder deals with the specified video rate. If you have 62, try your tests with that and see if I am still wrong. If I am I'll gladly retract my post. All I know is that if I cut my frame rate in half (with 62) I get aprox half the output (audio remains the same of course).

B0b
9th July 2004, 01:13
With VP 3.1, the fps does not matter.

Species8472
9th July 2004, 01:16
Well I used the vp6.2 with a few different settings and the On2 release of the nsv encoder (if it matters) for my test. And I got the same results as Marsman2k. Though I did use the originally released encoder and got the type of results Ken52787 had. Unfortunatly installing the On2 version erased the original version's settings so I couldn't do a test with my original settings that I got the Ken type results.

Marsman2k
9th July 2004, 02:46
Yes, glad you mentioned that 8472 - I recently installed the On2 nsv plugin as well. I had been using the latest release of the vp62 encoder .dll and I thought the same formula applied there as well. I've been working on rates between 100-150 and looking to get the best realtime video quality - thus have been leaning on the 62 side. My 61 work has been re-encoding other video formats to .nsv and been mostly 30fps anyway so I hadn't noticed that the 61 encoder maintained rate regardless of fps. Apologies for not bringing forth all the information in my original post.

PS. My original post is too old to edit - if an admin sees this perhaps they'd put a disclaimer at the top of the post indicating this applies to vp62 only.

Inedible Bulk
9th July 2004, 02:55
One may note that theres a good reason to stick with 6.1

There's no activex for 6.2

but then theres a good reason to stick with vp3 (or vp5, if someone were to get an encoder out)...

There's NSVPlayMoz support, as well as activex. I'd rather the stations run vp5 so people dont have to run IE to watch from inbrowser.

Jeeper One
9th July 2004, 08:39
Hi everyone:

There's no ActiveX for 6.2 yet BUT I've decided to put my cam on 6.2 anyway when I'm not doing my talk show as this gives me pretty much what I wanted my cam stream to be all along. :)

When I'm doing my talk show, it will be using 6.1 and will be on the MAIN BROADCAST page as that's where I'm going to move the show to anyway. :)

As a result of the lack of an ActiveX control viewer for 6.2 I have intentionally redirected the ActiveX control viewer on my cam page to show my vid stream. The external viewer link on the cam page should still work though if you have the VP6.2 decoder codec. :)

Plus, my vids are in 6.1 and this version does what I want for my vids.

Cam is usually 64K video with 16K MP3 audio. End result is an average bitrate of about 20 Kbps.

Vids are 32K video with 64K MP3 audio. End result is an average bitrate of 96 to 100K

Just my \$,02 worth :)

Cheers for now :)