I'm not sure in what forum to post this question, but I guess this is the most appropiate one... I live in Europe, where the TV standard is PAL at 25fps, but most of the movies I download are in NTSC framerates. Most DVD-players in europe play both PAL and NTSC movies. The question I'm struggling with is; should I convert my movies to 25fps progressive or not? I read this article on divx.com, and I think I understand the whole 3:2 pulldown and interlaced video thing (but I might very well be way off), so here are my thoughts so far: http://www.divx.com/support/guides/guide.php?gid=10 I’m afraid my knowledge of progressive video is very, very limited. Please, please correct me if I am wrong about any of this (and I probably am). 'Film' to PAL can also be done by slightly increasing the speed of playback (1 frame more per second, not noticeable). This is not a ‘real’ frame rate conversion (it’s not creating any new frames, like Canopus Procoder does), But the quality of video stays intact (until you code it to MPEG2). One app I found wich does this is EO Video. Most importantly, it alters the audio stream, so A/V sync stays intact. But what if the source is 29.9fps… Did the encoder of the Xvid or Divx set the codec to preserve the interlace (and 3:2 info)? Could I simply remove the pulldown info, and get a 23.98fps movie? I know Divx can be set to preserve this info, but how do I tell? And how do I remove it? Once I’ve got a FILM frame rate, and want to create a NTSC DVD, 3:2 pulldown info has to be added to create a interlaced DVD. 3:2 pulldown is the cause of NTSC DVD’s jerky playback in scenes that are supposed to have a smooth and slow horizontal motion. But what happens when I encode the DVD progressive?? Interlaced PAL uses 2:2 pulldown, which plays back smoother. No frames need to be ‘created’, every frame is simply divided into two fields. As appose to 3:2 pulldown, where every 2 frames equal 5 fields. This is how I create DVD’s at this time: 1. I let EO video render the MPEG4 source, at 25fps, saving both video and audio uncompressed (which can take up to 85GB of space). MOTIVATION: a. Adobe premiere doesn’t accept compressed audio. b. Trying this with Virtual- or Nandub will cause issues when the audio is VBR Mp3 this (even if I let premiere handle the frame rate conversion). c. Separating the audio with –Dub will cause A/V sync issues. Because I’m coding to PAL, and even if I where not, VBR Mp3 would still cause sync issues. 2. I import the huge file into Premiere pro. I let the build in MAIN-CONCEPT MPEG codec handle the conversion to MPEG2 DVD. I set the standard to PAL, two passes and progressive. I let the surcode AC3 plugin render Ac3 stereo audio at 192kbps MOTIVATION: a. I like Premiere. It’s easy to encode to MPEG fully compliant with official DVD specifications, and ensures the file is accepted by Adobe Encore. b. AC3 stereo is less lossy than MP2, or so I’m told. c. The above makes me think I’m best off encoding PAL progressive (????) 3. I author the DVD using Encore DVD. MOTIVATION: a. Encore allows me to easily add subtitles, which is a must. Again, I simply like it. I realise this is a lengthy post, but I’m trying to make you guys understand exactly what I’m struggling with. Please advise. Sincerely, John.
I'm far from an expert on conversion, but I'll see if I can help you out. Generally speaking, it's better to stick with NTSC as long as your player can read it, and obviously most PAL players and TVs don't have a problem with it. The problem with converting is that you'll always lose some quality from changing the resolution and re-encoding. This is correct. You can also do it with AviSynth, but you'd either have to use a third party program to use AviSynth with Premiere or open the AviSynth script with VirtualDub and save it as an AVI. There are 2 different reasons that a video file might be encoded at 29.97fps. Either the pulldown wasn't removed or the original was encoded at 29.97fps. If the original was encoded at that framerate, there are 2 possible reasons, and depending on which one it was you may or may not be able to convert with good results. The first possiblity is that the original was encoded with a hard telecine. That means that it comes from a film source, but was encoded with actual duplicated fields instead of adding pulldown to make your player duplicate the fields during playback. If that's the case, you could use AviSynth's Telecide and Decimate plugins to recover the original progressive frames, also known as IVTC (Inverse Telecine). It's also possible that the original came from a video or hybrid (mixed film and video) source. If that's the case, an IVTC operation will make any video content choppy because you'll be removing unique fields. Converting actual NTSC video sources to PAL is a difficult operation, and without spending thousands of dollars, you're not likely to get very good results. You might be able to do a decent job with Canopus Procoder, but since I've never tried it I can't really say. If you're converting to PAL all you should need to do is speed up the video and audio to the correct speed and resize the video to 720x576. I'm not familiar with EO Video, but it's possible to do the conversion with AviSynth and use a third party program to frameserve to Premiere. AviSynth will provide uncompressed audio without needing to save an intermediary file first (although you can use VirtualDub to save it as an uncompressed or HuffYUV encoded AVI to make things easier) I believe there's a plugin you can download that will read MP3 audio. Since the audio output from AviSynth is uncompressed, this should solve Premiere's problem with compressed audio. Using a couple of filters you can do the conversion pretty easily by using the following lines in your script: AssumeFPS(25,true) #Speed up video and keep audio in sync LanczosResize(720,576) #Resize to PAL resolution I know this is a confusing subject, and my explanation may not help much. If you want to try it, I can try to walk you through the process. All the software I mentioned is free, so you don't have anthing to lose trying it except maybe your sanity, but if you've been researching the subject you probably realize that already _X_X_X_X_X_[small]Looks like I picked the wrong week to stop sniffing glue DVD Rebuilder Guides: http://www.afterdawn.com/guides/archive/dvd_rebuilder_tutorial.cfm http://www.afterdawn.com/guides/archive/dvd_rebuilder_tutorial_advanced.cfm[/small]
Thank you so much for your reply I have made some progress. I now use virtaldub to change the frame rate to 25fps, and use Besweet to alter and uncompress the video (A/V synced). Instead of saving the whole thing uncompressed with EO Video. Oh, and Premiere already handles the resizing (I tried resizing to 720x576 in virtaldub once, but for some perculiar - Adobe - reason, it still needed enlarging in Premiere). I did try Procoder, and it does a very smooth conversion, but the quality decreases very very much, even when saving uncompressed video. So I guess that I should forget about converting to PAL if the MPEG4's source was a progressive NTSC DVD. Because then IVTC is impossible, and removing frames will lead to jerky playback, right? But how do I find out wether this is the case, other than simply trying and reviewing the result? Also, I have never used Avisynth before. But I guess I can learn. I managed php, I should certainly be able to manage these Avisynthscripts, right? If you could get me started, as you so kindly offered, please..
No problem. AviSynth is actually quite easy if you've done pretty much any kind of scripting. For basic scripts (like you'd be using) it's mostly a matter of understanding the syntax: Filter(arg1,arg2) Operations are done with filters. Some filters have arguments. For example, resizing filters take arguments for the resolution: LanczosResize(720,576) Some filters don't take any arguments and some arguments are optional. You can add more filters by copying plugins (DLL files) to your AviSynth plugins directory (C:\Program Files\AviSynth 2.5\plugins\). To IVTC film sources you'll need the Decomb plugin. You can get that, plus many others, from http://www.avisynth.org/warpenterprises/ AviSynth is available from http://sourceforge.net/project/showfiles.php?group_id=57023 You'll want either version 2.54 or 2.55 Here are the lines for a very basic script to convert an MPEG-4 AVI with MP3 audio from 23.976fps NTSC to 25fps PAL: AviSource("C:\Wherever\filename.avi") AssumeFPS(25,true) LanczosResize(720,480) Using a 3rd party program (not sure what it's called right now) you can open that script in Premiere and do whatever you want with it. I don't know how AviSynth's resizing compares with Premiere's so you might want to make a comparison for yourself. If the video is 29.97fps, but the source is film, you should be able to IVTC it to recover the original film source. This is the same thing an NTSC progressive DVD player normally works. To do that, change the script to this: AviSource("C:\Wherever\filename.avi") Telecide(order=1,guide=1) Decimate() AssumeFPS(25,true) LanczosResize(720,480) If the source was recorded on video tape instead of film, you're pretty much out of luck for high quality conversion. You can find my AviSynth guide at http://www.afterdawn.com/guides/archive/avisynth_tutorial.cfm
This I don't understand: The rest is pretty much clear. I just have to find out what that 3rd party app is called. Thanks a BUNCH. P.S. This won't work though, but you know that AssumeFPS([bold]25[/bold],true) LanczosResize([bold]720,480[/bold])
You did, I already downed it. Thanks again. If you happen to recall what that app(or plugin) for Premiere is called, please let me know. I still don't fully understand this, I'm afraid. I'll figure it out. You've been a lot of help, thank you. (I'm gonna stop thanking you now... )
Here's the plugin for Premiere: http://sourceforge.net/projects/videoeditorskit/ If you do a Google search for 3:2 pulldown or 2:3 pulldown you should find some pages that explain it a lot better than I could, but I'll try anyway. Basically the way it works is duplicating fields (half frames) to change the framerate from 23.976fps to 29.97fps. The normal field order would look like this: Field A: 1 | 2 | 3 | 4 Field B: 1 | 2 | 3 | 4 With pulldown flags added, the video stream wouldn't actually change, but a player would read it like this: Field A: 1 | 1 | 2 | 3 | 4 Field B: 1 | 2 | 3 | 3 | 4 By doing that, 4 frames become 5 frames for display purposes only. However, sometimes video is actually encoded with that pattern, which is what I referred to previously as a hard telecine. With a hard telecine there are no pulldown flags, so you have to use software that can recognize the pattern in actual interlaced video. That's what Telecide() does. Decimate() then removes the extra information for you.
So what you're saying is: there are DVD's coded with hard telecine, and there are dvd players which are able to do IVTC. But then it plays at ~24fps??? Or does the DVD-player render progressive frames by itself? ...... for the link to the plugin!
The video will always be read as interlaced initially because there's no such thing as progressive video in the DVD specs. After that, most of them can IVTC the video. In reality, 2:2 pulldown isn't pulldown at all (since the number and placement of fields is the same as the original). It's actually just a workaround to be able to distinguish progressive PAL video from interlaced. In reality, different progressive scan players can do different things, depending on the deinterlacing chip used. Some will only read flags, meaning that a hard telecine will be ignored and deinterlacing will be done wrong on the video using some combination of interpolating (a fancy term for an educated guess) and blending fields. Others only use pattern matching, meaning that a correct telecine should always be detected and a hard telecine will usually be detected. Some only do interpolating and blending, meaning you'll get pure crap from telecined video (whether it's through pulldown flags or a hard telecine) because no IVTC will be done at all. The best deinterlacers use adaptive deinterlacing, meaning that if pulldown flags are present IVTC will be performed, but if they're not present the deinterlacer will attempt to match fields to see if it can be IVTCed. If it can't match the fields, it will revert to interpolating and blending. One of the biggest problems with most deinterlacing chips (at least the ones used in player sold in the US) is that they don't recognize 2:2 pulldown, even if they do PAL to NTSC conversion. I suppose it makes as much sense as anything else in the world of DVD - none at all.
Just one more question: What do you think the MainConcept codec in premiere does when it lets you pick progressive encoding? Because you said there's no progressive in the DVD specs... Í'm getting more and more confused.. That was hardly a joke...
When you select progressive encoding, you'll get progressive video. That's why the pulldown flags exist, so progressive sources can be pretend to be interlaced. And you're right - the sanity remark wasn't really a joke. I've spent so many hours reading and re-reading (and re-reading and re-reading) different sources that if I wasn't so stubborn I would have given up long before I understood this.
Just to check if I understand: the vid is progressive, pulldown info is added, but [bold]ignored[/bold] by the player? I think I need to forget what I think I know and read a lot a lot a lot of info... Got any good sources?
The pulldown flags are used to make it read as interlaced (ie one field at a time instead of a whole frame at a time). In the case of a good progressive scan player, it can be used to combine interlaced fields properly (and decimate 1 out of every 5 frames for telecined video) to re-create the original progressive frames. In other words, when 2:2 pulldown flags are read, the top and bottom fields from each frame are simply combined together to be sent to a progressive scan TV at the same time. Otherwise the video would be sent to the TV one field at a time, meaning the even lines would be drawn first, and then the odd lines would be drawn next. For NTSC video that's actually telecined 23.976fps progressive with pulldown flags, a progressive scan player will start by combining fields that were originally in the same frame together, which will leave a stream where every 5th frame isn't actually from the original stream. Then the extra frames are removed, and the stream is played at 23.976fps instead of 29.97fps. A regular DVD player will simply read the fields specified by the 3:2 pulldown, meaning some fields are displayed twice. I wish I knew of a good source for this information, but I mostly spent many hours using Google, and reading the same information over and over again from different sources. A lot of it comes from support sites for different software packages, so you often have to separate instructions for specific software from general information about the subject. My recommendation is to start by making sure you have a good understanding of how interlaced video works. Without that it's hard to understand what pulldown flags do. After that, I'd Google for "pulldown", "3:2 pulldown", "2:3 pulldown" (don't ask - I don't want to confuse you more), "telecine", and "IVTC". _X_X_X_X_X_[small]Looks like I picked the wrong week to stop sniffing glue DVD Rebuilder Guides: http://www.afterdawn.com/guides/archive/dvd_rebuilder_tutorial.cfm http://www.afterdawn.com/guides/archive/dvd_rebuilder_tutorial_advanced.cfm[/small]
