How do I use avisynth to frameserve multiple movies?

[sneaux]

I read about a guy who framserved 26 episodes of anime to make a music video. I'm trying to do something similar to that. He said he used a vfapi reader and avisynth. I have both of these programs but I'm not sure how to use them. I get the scripting part of avisynth, but what do I do with these scripts once I'm done? What opens them?

[manono]

Hi-

What opens them?

VDub(Mod), CCE, TMPGEnc, and many other encoders.

[kschang]

The question is... WHAT exactly are you trying to accomplish here? What is the end result you wish to achieve? You've describe the parts, but not the expected end result.

Do you want to burn them all to a DVD viewable on standalone DVD player?

Do you want to just archive them as easily viewable data on a DVD-R?

WHAT are your intentions?

[sneaux]

I found a way to use the Half Life 2 demo recorder to export targa (.tga) sequences of the game to make a high quality movie. My intention is to put all the 8 hours or so of gameplay on a few dvds. I need to be able to edit so I'm opening the targe sequence in virtualdub>open VDR file with Link2>then open dummy avi file in premiere. I don't want to compress to anything like DV because quality goes down.
The problem is I am going to have image sequences from several dozen sections of the game. The way I'm doing it now I would need to open several dozen dummy avi files which would also require several dozen instances of link2 and Virtualdub. Do you a way I could open several dozen dummy avi files served from a single frameserver?

[guns1inger]

So you eventually want to create a complaint DVD of this footage (as many disks as required, obviously) while retaining quality.

Why do you need Link2, when pretty much any encoder of quality can read virtualdub frameserver or avisynth files directly ?

What encoder do you intend to use ?

You can load avisynth files from within avisynth scripts. You could create a set of scripts for different sections of the game, then one script that loads each of these using the Import command.

Or you could just chain ImageSource() + ImageSource() + ImageSource() where each call to imagesource opens a new section of images.

[sneaux]

I think those avisynth solutions would be perfect but I was wondering if you could explain how to make the scripts more clearly for me. I just discovered how to use the virtualdub frameserver a week ago and today was my first time actually getting avisynth to do something (opened a wmv file in tmpenc.)
I was using link2 because virtualdub makes a vdr file which wax, vegas, and a lot of other programs can open but not premiere. Premiere can only open the dummy avi file link2 makes. I will be happy to switch to avisynth because link2 wants me to pay $15 for it. I'm just an avisnth noob so I would be very grateful to get the long version of those two solutions you mentioned. How do I link image sources together and use the import command?

[manono]

Here's some information on how to use the ImageSource command:

http://www.avisynth.org/ImageSource

[sneaux]

Heres something that would have solved all my problems from the begining. I was in dv mode of premiere and when I go to windows avi there is a frame option in there and I can open all my targa sequences in there.

That solves my frame erving dilemma though I still have some questions I hope someone can answer. I'm using the debugmode frameserver to export my premiere edits to tmpgenc and debugmode asks if I want to run it in RGB24. RGB32, or YUV. Which one is the best for uncompressed targas?

Also, the targa frames are square pixels and I was wondering if I could keep that pixel dimension in the mpeg on the DVD.

Last thing. These uncompressed frames are still huge. a 1280 by 720 frame is 2.7 megs. I've heard about mjpeg and I was wondering if it does what it sounds like it does. Would it convert the targas to jpegs and make is a movie. How do I convert to mjpeg?

[mikiem]

"asks if I want to run it in RGB24. RGB32, or YUV. Which one is the best for uncompressed targas? "
"targa frames are square pixels and I was wondering if I could keep that pixel dimension in the mpeg on the DVD."
"These uncompressed frames are still huge. a 1280 by 720 frame is 2.7 megs. I've heard about mjpeg and I was wondering if it does what it sounds like it does. Would it convert the targas to jpegs and make is a movie. How do I convert to mjpeg?"

OK, first off the DVD spec is a non-square pixel, non-rgb, 720 x 480 or 576 frame, so basically everything has to change. Mjpeg or mp4 or wmv etc don't have the pixel aspect or frame size constrictions, though on a DVD would have limited compatibility. Mjpeg would produce files that are huge, but is the easiest done -- simply render from prem to mjpeg avi.

Colorspace I think would depend on the software you were sending the video to, but I would suggest YUV, the basic type used by video.

For DVD:
Pixel aspect needs to be changed for DVD mpg2, since any playback device or software is going to assume it's non-square and apply a correction. There are a few ways to accomplish this, but I recommend following the docs on recommended still sizes in Adobe's Knowledgebase. Alternatively you might find it practical to resize in V/Dub, which will be magnitudes faster. Doing the resize, you'll need to either crop or add letterboxing, & go anamorphic.

[manono]

Hi-

TMPGEnc uses RGB24 input, is going to convert it if the source isn't RGB24, so you'd best do it before it hits TMPGEnc.

If you allow TMPGEnc to do the resizing to 720x480/576, it'll use crappy Bilinear resizing on it, so you'd best resize it before it gets to TMPGEnc, using a better quality resizer.

Since your source is 1280x720 1:1 square pixels, all you have to do is to resize directly to 720x480/576 and make sure you encode for 16:9. No additional cropping or letterboxing is needed.

[sneaux]

Okay I will use rgb24 when frameserving with debugmode.

I also noticed I can export to avisynth in premiere and open the avs file with tmpgenc. It gives me the code IPCSource("1280by760.") 1280by760 is just the name I gave to the frameserver.

Could I add more script to the avs file to convert to rgb24 and resize to 720 by 480 before it gets to tmpgenc?

Also I wanted to get your opinions on what resolution to capture in HL2 for best quality. There are six resolutions for 4:3 but I don't want to encode in that format. It can play 16:9 at 848*480 and 1280*720 and 16:10 at 720*480 and 1280*768.

What is 16:10? I've never heard of that aspect ratio before. I want to use it because it is already the perfect dvd size but what will happen when I try to encode a 16:10 picture to a 16:9 dvd? Will this be better or worse than resizing 16:9 848*480 to 720*480?

[manono]

Hi-

For the color conversion and resizing, I'd use:

LanczosResize(720,480)
ConvertToRGB24()

along with your IPCSource line at the top. Before sending it to TMPGEnc, I'd first test the script by opening it in VDubMod to see if it's OK, and if not, what the error message is.

If 1:1 also, 848x480 might be better, because then you'll only be resizing along 1 axis, and the encoding will also go much faster.

I don't understand what 16:10 means in that context either. If it's still 1:1 (square pixels), then it won't do you much good. However 720x480 might be useful if it's not square pixels already. You might capture a few frames at that resolution, stretch it out to 848x480 and see if it looks "normal". If so, then no resizing at all will be necessary, and that would be the way to go. I doubt it's like that though.

[sneaux]

I'm having trouble with the avisynth premiere plugin. It gives me the code IPCSource("843by480.avi"). Premiere gives me a dialogue box allowing me to save 843by480.avi on my desktop but nothing ends up on the desktop. I thought maybe that was normal so made an avs file on the desktop with IPCSource("843by480.avi") and tried to open it in vdub but there it can't find the file. Anyone know how to use the IPCSource command?

[manono]

Why don't you see if AviSynth's ImageSource can open your TGA files directly:

http://www.avisynth.org/ImageSource

It doesn't seem to say one way or the other.

I've never used Premiere, but the IPCSource command seems to have been discontinued:

http://www.avisynth.org/IPCSource

Maybe if you install old versions of AviSynth you can get it working. Looks like the DebugMode FrameServer is the way to go now (which you know already):

http://www.debugmode.com/frameserver/

[sneaux]

I was able to make a script that opens the sequence as tga and jpg after converting with the photoshot image processor. To save space I was thinking about converting every frame to jpg. Can I do that with the imagewriter in avisynth. So far all I've been able to make it do is export as ebmp. I think if I can get this to work it will be far faster than photoshop.

I had some questions about colorspace. Does colorspace apply to still frames. What colorspace is tga and jpg? Some people have told me I can keep quality by staying in the same colorspace but how do I find what colorspace is a tga frame.

Here is a singe tga frame if you want to see what I'm talking about 1_10313.tga
If you decide to look at the frame I'd appreciate suggestions on how best to preserve the aspect ratio

Lastly-I'm having trouble getting the sound to play along with the image sequence. I put in
ImageSource(file="D:\Half Life\848 480\1_%05d.tga",end=12934,fps=29.97)
WavSource("D:\Half Life\848 480\1_.wav")

but then it only play the audio


Thanks for the help guys

[manono]

Hi-

A lot of questions, most of which I can either only guess at or can't answer at all.

As for creating JPGs, I think you'll get only what the encoder can encode to. So, VDub will make you an AVI, and CCE an MPEG-2. That one's just a guess. There are probably batch converters in some of the Image Editors, but I don't do any of that. I'll bet a lot of people around here do, though.

I have no idea what colorspace your TGAs are in, but I know how to find out. Take the .avs with your ImageSource line for the TGAs and add:

Info()

at the bottom, save it, and open it in VDub(Mod). It'll give you the colorspace of the file and a bunch of other information as well.

You're NTSC, right?? You'll keep the Aspect ratio (for MPEG-2) by resizing to 720x480 and encoding for 16:9.

VDub and its varients aren't really designed to be players. I wouldn't even include the audio in the script if I were you. But maybe that's just me. Oh, you didn't say what you were using to try and play it. I don't think I'd worry about it though. You see your pictures when you open that script in VDub and scroll around, don't you?

[sneaux]

Thanks for the info() idea. I now know that the frames are rgb24 which is good because I can use tmpgenc without having to worry about chaning colorspace.

I thought of another question about colorspace. So the frames are already rgb24. When I export them with the debugmode frame server in premiere I have the option to choose to serve as rgb24, then tmpenc can only work in rgb24 then apparently in the last step to mpg it is converted to yuv. My question is what happens when you convert to rgb24 over and over again. Does debugmode and tmpenc know it's rgb24 already and doesn't actually change anything, or does it convert to rgb24 again. If so does that process remove quality in any way?

[manono]

Yes, all colorspace conversions degrade the video a little bit, I think. But you're safe, I think, as it's staying in RGB24 all the way through until the end when TMPGEnc spits out a YV12 MPEG-2. So no, TMPGEnc isn't doing any conversion on it if it's already RGB24 when being input to TMPGEnc.

Like you though, colorspace isn't really my thing. I've learned the same thing you were told, do as few colorspace conversions as possible, filter in the original colorspace if possible, and if using AviSynth, do any necessary conversions in AviSynth before sending it to the encoder (YUY2 for CCE and RGB24 for TMPGEnc). So someone can feel free to correct me or to elaborate.

[sneaux]

I was trying to open several videos in the same script and was having trouble with small differences in resolution and audio sample rate. I can't link clips together when those things are different.

AVISource(filename)+
AVISource(filename)
Lanczos4Resize(target_width,target_height)+
AVISource(filename)+
AVISource(filename)
ResampleAudio(new_sample_rate)+
AVISource(filename)

I tried writing a script like this but it doesn't work. The only way I can link different clips is by making another script that resizes then open the avs in the link-everything-together-scrpit. Is there a way I can link files of different audio saples and resolutions in a single script?

[manono]

Hi-

I don't do audio in the script, so I can't help you there. However, after you see some examples, you may be able to adapt it with the audio included. And it's not even necessary to join them before encoding anyway. You can encode them separately and then join them when authoring the DVD. For joining AVIs (but with the audio disabled):

A=AviSource("AVI1.avi", false)
B=AviSource("AVI2.avi", false)
C=AviSource("AVI3.avi", false)
A=A.Lanczos4Resize(xxx,xxx)
B=B.Lanczos4Resize(xxx,xxx)
A+B+C

Something like that. Open the script in VDub(Mod) to get specific error messages.

[sneaux]

Thanks, that code worked. I was wondering why you didn't like to work on audio in avisynth. Are there disadvantages?

I want to normalize each separate channel in each clip too. This is what I have so far. I know it's not quite right because the audio comes out mono and I know it's a stereo source.


A=AVISource("D:\Amazon\1.avi")
B=AVISource("D:\Amazon\2.avi")
C=AVISource("D:\Amazon\3.avi")
D=AVISource("D:\Amazon\4.avi")
E=AVISource("D:\Amazon\5.avi")
F=AVISource("D:\Amazon\6.avi")
A=A.Lanczos4Resize(352,240).ResampleAudio(48000).G etLeftChannel().Normalize.GetRightChannel().Normal ize
B=B.Lanczos4Resize(352,240).GetLeftChannel().Norma lize.GetRightChannel().Normalize
C=C.Lanczos4Resize(352,240).ResampleAudio(48000).G etLeftChannel().Normalize.GetRightChannel().Normal ize
D=D.Lanczos4Resize(352,240).ResampleAudio(48000).G etLeftChannel().Normalize.GetRightChannel().Normal ize
E=E.Lanczos4Resize(352,240).ResampleAudio(48000).G etLeftChannel().Normalize.GetRightChannel().Normal ize
F=F.Lanczos4Resize(352,240).ResampleAudio(48000).G etLeftChannel().Normalize.GetRightChannel().Normal ize
A+B+C+D+E+F

I've been experimenting with MergeChannels and GetAudio but I don't know exactly where to put the script. It keeps saying syntax error.

[kschang]

Just personal preferance, but there's no need to

A=A.Lanczos4Resize(352,240)

Just do

A.LanczosResize(352,240)

And also, if you're SHRINKING the video, there's no need to use Lanczos. Use Bicubic or even Bilinear will be MUCH faster.

Personally, I would separate the audio and process them separately with a tool like Audacity, then mix it back in. However, if you wish to play that way...

left_ch = GetChannel(a,1).Normalize
right_ch = GetChannel(a,2).Normalize
audio = MergeChannels(left_ch, right_ch)
AudioDub(a, audio)

I suggest getting AVSEdit, if only for the helpfile that came with it and the examples.

[manono]

Hi-

Is that the first time we've seen the final resolution? I agree with kschang that Lanczos4Resize isn't the best solution. Ordinarily I prefer regular LanczosResize, but at that low of a resolution I don't guess it'll help much and Bicubic will certainly be a bit faster. Personally, I steer clear of Bilinear resizing at all costs.

I was wondering why you didn't like to work on audio in avisynth. Are there disadvantages?

None that I know of. I don't do much audio work requiring the kinds of things you're doing, and have never bothered to learn how to do it in AviSynth. I use a WAV Editor, such as CoolEdit Pro. So I wasn't being disapproving; I just didn't know how to modify the scripts including the audio. But kschang does.

[kschang]

I just read the examples. Really.

[sneaux]

Thanks manono and kschang,

With your help I was able to get get those individual channels normalized.
I just started using avisynth a month ago and I love it.

I guess generally bicubic is best for lowering resolution and LanczosResize and Lanczos4Resize are better for raising resolution. I will keep that in mind.

One more question and I think I will have run out of questions to ask... I can't open avs or vdr files in Premiere. Are there plugins I need to get?

[kschang]

I think there's a Premiere plugin you have to download. Search the guides or the forums, it's listed somewhere in there.

[manono]

Yes, you'll need the plugin, and maybe HuffYUY if you don't have it. You can get the plugin lots of places I'm sure, but here's one:

http://videoeditorskit.sourceforge.net/

And here's a page, although old, that will help get you going, if you have other problems:

http://www.animemusicvideos.org/guides/ErMaC/guide5_2.htm