BTW (someone correct me if I'm wrong) but in my experience "Psychovisual enhancements" makes not one bit of difference in the size of the output file?
| Printable Version of Topic
Click here to view this topic in its original format |
| Unofficial VirtualDub Support Forums > Advanced Video Processing > Highly Compressible Tv Show |
| Posted by: jcsston Nov 18 2002, 05:16 AM |
| I have some that I think is a bit funny. A local TV station broadcasts Star Trek Voyager reruns Late Sunday Nights. I programmed my computer to record them. Now whenever I compress them using DivX my average bitrate is around 480kbit no matter how high the bitrate is set even on two pass. Now at first I thought my filters was doing this but even with different filters my output size is the same 150MB, even for different episodes. In DivX I have all the Mpeg4 Enhancements enabled and the Pycovisual to Strong. When I watch the compressed video it look okay to me. I use ATI TV to record them to MPEG-1 320x240 at 8Mbit could this cause them to compress so well? |
| Posted by: fccHandler Nov 18 2002, 06:23 AM |
| IMHO, 320 x 240 is "lo-res", but hey, whatever works for you. If you're happy with it, then we are all happy for you! BTW (someone correct me if I'm wrong) but in my experience "Psychovisual enhancements" makes not one bit of difference in the size of the output file? |
| Posted by: BaronVlad Nov 18 2002, 09:31 AM | ||
Hi fccHandler, I think you are right, with both  320*240 is very low and your problem should be because of MPEG1, I had many movies in MPEG1 that i wanted to compress with DivX into 700 MB. No Chance, they always were about ~400 MB in the end, no matter, what bitrate I had. I also made the experience, that there was no difference in filesize between "Psycho"-movies and "Non-Psycho"-movies. You should capture directly with VDub (Codecs Huffyuv or MJPEG) at a higher resolution (Full PAL/NTSC if possible) and then (resize to a smaller resolution !!! and) compress into DivX, it is better quality. Or you can try MPEG2 recording (i.g. using PowerVCR) and then let GKnot do the compression for you. Or you are just happy, that your files are so small that you can burn them easily an a normal CD-R |
| Posted by: jcsston Nov 18 2002, 09:16 PM |
| Well I also tried capturing at 640x240 MPEG-2 and the end result was the same. BTW Capturing with VirtualDub isn't an option since the show comes on at 3:00AM. VDub VCR doesn't work with my Ati Aiw. |
| Posted by: BaronVlad Nov 19 2002, 08:05 AM | ||
??? Then your ATI AIW is the only ATI AIW that doesnt work with VDub VCR, normally it works wonderful. |
| Posted by: jcsston Nov 19 2002, 07:15 PM |
| Well it works with the timer but I can't get it to change channels automatically. My processor is an T-Bird 650Mhz and it cannot play 640x240 DivX that well and that's why I'm stuck with the miserly low resolution of 320x240. But it's better that my previous setup 250mhz 15fps  |
| Posted by: jcsston Nov 19 2002, 10:30 PM |
| After reading about it online I finally downloaded the fddshow filter. It can play my DivX 640x240 smooth! |
| Posted by: jcsston Nov 25 2002, 01:47 AM |
Well this Sunday I got another episode and this time I captured it at 640x240 MPEG-2 and compressed it at 640x240. Guess what the file size was at the max bitrate. 312MB  And it looks a good bit better than before  and I can still fit about 3 on cd. (at a bitrate slightly lower than the max)
|
| Posted by: Spire Nov 25 2002, 01:53 AM | ||
If you're going to increase the resolution on one axis, you're much better off doubling the vertical resolution; e.g., 320x480. This will preserve both fields of the original video, resulting in much better motion. |
| Posted by: jcsston Nov 25 2002, 02:02 AM |
| Wow that was a fast reply. Anyway in the ATI MPEG-2 setup should I enable or disable the deinterlacing? |
| Posted by: Spire Nov 25 2002, 02:05 AM |
| If you're capturing with a vertical resolution of 480 (which I recommend), you should definitely disable deinterlacing. |
| Posted by: jcsston Nov 25 2002, 02:23 AM |
| Thanks I'll try 320x480 with the Smart Denterlace filter in VDub. |
| Posted by: fccHandler Nov 25 2002, 03:55 AM |
| Even today, many television shows (and Voyager too, probably) are shot on film, so for best results I recommend under Video->Frame Rate, click Inverse Telecine "Reconstruct from fields - adaptive", and set Smart Deinterlace to "blend instead of interpolate" for when the IVTC fails. Motion will be smoother, image will be sharper, and you'll get better compression with the frame rate reduced by 4/5. Hope this helps. |
| Posted by: outlyer Dec 11 2002, 04:05 PM | ||
Although thread has went a bit off-topic  I'll give you my 2 cents on the original question: Modern codecs, like DivX, usually respect you'r bitrate settings BUT if they're too agressive they'll be ignored: if a frame needs only X bits to get a "100%" quality, the codec wont give it more than X bits.You can get an appropiate bitarte with a bitrate calculator, in this case, doing some rough numbers and using gordian knot it gives a high objective quality (based on dimensions, duration and bitrate), so it fits to what I said above |
| Posted by: S_O Dec 11 2002, 05:31 PM | ||
I record sometimes from TV, too, but in PAL 25fps, is there any way to reconstruct the original 24fps here? Iīve read they use speed-up, the film plays faster than normal, but anyway there is interlacing, how does the interlacing come into the film in PAL? Is it reversibly? |
| Posted by: outlyer Dec 11 2002, 07:02 PM | ||
In PAL there's no such thing like telecine, interlacing is simply done, AFAIK, splitting every frame in two fields (with alternative lines, you know). As you mentioned they speed up films to convert 24fps to 25fps (audio included, so original audio has a bit of a lower pitch), so there's no frame repetition or any weird thing used in telecine, (really straight: 24 frames per second => speed up => 25 frames per second => 50 fields per second). I've done some reconstructions from DVDs, simply slowing down and changing the audio pitch to match the duration, and I've got good results. |
| Posted by: Morsa Dec 11 2002, 07:06 PM |
| Sometimes the audio speed up is performed without changing the pitch. Also sometimes they really add an extra frame to maintain the same duration. |
| Posted by: jcsston Dec 11 2002, 07:09 PM |
Is telecine possible with 240 height video? The telecine setting under framerate options causes my video to have interlacing artifacts  |
| Posted by: outlyer Dec 11 2002, 07:20 PM | ||
@Morsa ok, so, usually "what I said" anyway... how can audio been speed up without changing the pitch? |
| Posted by: fccHandler Dec 11 2002, 07:26 PM | ||
You need to capture at height = 480 to get both fields. |
| Posted by: S_O Dec 11 2002, 07:50 PM | ||||||
frame A is splitted into field A1 (line 1, 3, 5,...) and A2 (line 2, 4, 6,..), same for frame B. But if I capture both fields again into one frame it would be without interlacing, so the fieild order must be something like that:
The two fields one above the other means one frame. That could be easily converted back, why doesnīt that filter excists?
|
| Posted by: fccHandler Dec 11 2002, 08:21 PM | ||||||
If that's the case, it's easily fixed with Avisynth:
|
| Posted by: Morsa Dec 12 2002, 07:17 AM |
| I use Cooledit, and it changes duration without changing the pitch. All professional Audio editors performs this task too. Believe me, I had worked with audio for cinema many times. In the digital domain almost everything is possible!!! |
| Posted by: outlyer Dec 12 2002, 01:00 PM | ||||
I believe you, it's only that I don't see how can audio been speed up without changing the pitch. It isn't such an important thing for me as it's hardly noticeable al least to me (by the way, I also use Cooledit to do this ).
I can't remember but I guess that they were non interlaced, as I was doing it like an experiment probably I choosed the easier films to do this... I haven't the knowledge to answer your questions about field ordering, I just knew how the fields are obtained; maybe in http://www.100fps.com/ there are some answers, I don't remember 
|
| Posted by: muf Dec 12 2002, 02:02 PM | ||
Three options here .1) Your capture card screws up fields in some way so the original progressive frames don't align with the combined-field progressive output. 2) Film is converted to NTSC and the NTSC is converted to PAL with generic NTSC conversion (no IVTC or anything)-- your video is screwed up: you may kick your broadcaster in the ass. 3) Star Trek ISN'T shot at 24fps . |
| Posted by: outlyer Dec 12 2002, 08:52 PM | ||
Oh! S**t I forgot to mention this , really there are some PAL films, series, etc which follow what muf said and if this is the case it will sure be a pain in the ass |
| Posted by: Spire Dec 13 2002, 08:47 AM | ||
It's done by resampling/interpolating the audio waveform. The same technique is also used to change the pitch without changing the speed. |
| Posted by: sh0dan Dec 23 2002, 11:54 AM | ||
Time-stretching is a very destructive way of changing the length of a sample, and I'd recommend against it. There are several ways of doing it (zero-cross / spectral split, etc.) but ALL of them are quite destructive. You loose "attack" and clearness of your sound, because there are small portions of the sound taken out, or repeated. Change the samplerate instead, and make it play at a different pitch - it'll give you a much better result. |
| Posted by: S_O Dec 23 2002, 01:45 PM | ||||
1.) ? 2.) I know they did it that way for the older series (star trek TOS & TNG), but for the new series (DS9 & VOY) I donīt thing they have done it in a that crappy way. 3.) Star Trek is 100% shot on film, so most likely on 24fps, or does someone shot on 29,97fps film?
Download this: http://www.sunpoint.net/~oparviai/soundtouch/ Itīs a open-source sound-processing-lib, which can in/de-crease playback-rate/tempo/pitch. There is also a winamp-plug-in (use the new beta, the quality is better) and itīs quite high-quality. Sometimes when you hear exactly at a old movie you can hear the crappy methodes, it sounds sometimes like xing-mp3, unsharp. |
| Posted by: jcsston Jan 13 2003, 06:29 AM | ||
It's a mixture  in the episode 'Caretaker Part 1' at the beginning with the text scrolling up it is fine no interlacing artifacts at all. But then two ships come on and start firing at each other with interlacing artifacts, every field is different.Once they switch to interior view of a ship 24fps or duplicate fields for the rest of the show  |
| Posted by: fccHandler Jan 15 2003, 07:55 PM |
| I have a passion for music videos, but they are the worst. Apparently the editing is done after the 24fps films have been telecined to 29.97 NTSC. As a result, the telecine pattern shifts at nearly every scene change! Even VirtualDub's automated IVTC chokes on these. Some interlaced frames always sneak through, and the output sometimes seems to jerk or skip, as I suppose it's trying to keep A/V sync. Still trying to find a perfect solution. |
| Posted by: Kippesoep Feb 1 2003, 08:56 PM |
| Many times, they seem to convert the telecined 30fps NTSC version to 25fps PAL causing some PAL frames to contain information from multiple frames of NTSC data. I hate that. Does bad things to motion detection. |
| Posted by: jcsston Feb 2 2003, 03:00 AM | ||
I drove myself crazy trying to manually do IVTC on CareTaker Part 2. I was using the beginning part where they review what happened on the previous episode. I forget that they probably edited the 24fps film after it had been telecined. There is a http://forum.doom9.org/showthread.php?s=&threadid=43362&highlight=star+trek over on Doom9 started by neuron2 (Donald Graft) about doing IVTC on Star Trek clips. |