View Full Version : AVI or AVI DV ?

03-30-2008, 09:35 AM
Hi all, i'm working to a large viz movie (10 min). I render PNGs then i encode sequences in uncompressed AVI format, and finally i use these for final editing in Premiere.
I'm getting very large files, so i tried AVI DV (pal) format; it's claimed to be a lossless compression but it's nearly 1/10 of uncompressed AVI size.
I'd like to know something more... is it safe to use AVI DV as base files, or should i stick to unc. AVI ?


03-30-2008, 10:33 AM
Well, someone with alot more video savvy than me will hopefully add to this.....

But.....we're delivering shots for a broadcast show at the moment, PAL standard def, and at the request of the editor we were uploading each shot as a PNG image sequence, with each frame creating a 1Mb or more file. Mounts up when your shots are 500 frames long. Now because of their setup at the edit suite, they've asked us to send shots as PAL DV Quicktimes, which is creating much smaller files than we were using.

The editor(s) haven't commented on quality, or compression problems, or indeed any kind of problem.....that's not a very techy answer, but hopefully it might help a bit.


03-30-2008, 11:04 AM
interesting, i'll give it a try. I was wondering if there's something like LZE lossless compression for TIFF, but for AVIs.
This project is eating 100s GB of HD, from old versions to new ones, and so on.


03-30-2008, 11:23 AM
I think as DV you get the files compressed, I think it is something with 1:5... this isnt a great choice for production-quality... loseless isnt that small size, but why you dont use QuckTime ? there you can save the .mov as a png-compressor, and png supports also Alpha :-) and it isnt bigger than all of your .png files :-)

I was a bit cynical about using DV compression to be honest.....I'll try png-compressor.....thanks!


04-01-2008, 11:36 PM
Strictly for the numbers game: DV is a 5 to 1 compression ratio. Roughly speaking this means 1/5th of the original information is kept. But that is actually deceiving. Modern compression is all about visually perceptive artifacts control. In other words, your not taking your original signal and throwing out every 5 pixels, for the one you keep (which would look insanely bad). Instead it uses Discrete Cosine Transform (a fancy math analysis of the pixels) to squeeze the original image "coming off the chips" or in the case of CG the render out from the software to a smaller amount of data, with as little visual impairment as possible. This codec is intraframe (compressing each frame individually), as opposed to an interframe (temporal) codec, like MPEG, which also performs intraframe compression.

Bottom Line: The biggest hit for me in day to day production when working with the DV codec is the color space compression, which is 4:1:1, meaning only a quarter of the color info is accurately recorded and this trully makes keying difficult. Another hit comes with edges, all sorts of funky things can happen. My advice, don't encode to a video stream/codec. You always run the risk of a crash wiping out all your render time and you are "sewing" your images into the fabric of the codec.

If space is an issue, I have successfully used JPEG at high quality, JPEG 2000, png, etc... and dropped it into the final DV project.

A thought to keep in mind: what is your finally delivery format and what are you going to do to the images before they arrive at output. If you aren't doing anything fancy to your renders (extensive color correction, transforms, layering, etc) you don't need the same level of resolution (not talking pixel dimensions, but colorspace, pixel integrity, aliasing) as something that is going through crazy levels of manipulation.

Every bit of imagery humans create is compressed somehow and at some level compared to real life. Lenses compress because they don't transmit every single bit of information from the real world. Apertures don't let all the light in. Frame sizes don't allow all the information in a scene to be captured, Film grain, chip pixels all these things "shrink" the amount of "data" Also, when you record at x number of frames per second you are filtering out a lot of information (wonder what the frame rate of reality is? hmmmm).

So, while "uncompressed" is always a hot buzzword, it isn't always necessary or even desirable (unless you like slow calculation, huge storage requirements and slower transfer rates). Compression, like everything else in our world, is just a tool. Wikipedia and many other resources on the net have all the info you could ever want on all the codecs, formats and compression methods so rock on and expand your toolset :)

Have Fun,
Kevin L

04-02-2008, 01:59 AM
Thanks for the infos, very exhaustive.
My issue is mainly space and speed of tranferring of "base sequences".
I render PNGs in LW, then i encode these in uncompressed AVIs for final editing.
I'd like to replace unc. AVI with something not that big, but DV's 1/5 ratio seems to be too compressed, honestly ;)
I've to say this comes from Premiere's issue: some things cant be done properly on image sequences, i.e. sequence stretching or some frame interpolations are done on pre-encoded sequences only.
If i hadnt this issue, i could easily skip this intermediate encoding and use directly my beloved sequences... any cheap but powerful editing software around which doesnt need pre-encoding at all ? I missed SpeedEdit offer and i still beat myself for this... ;(


04-02-2008, 03:13 AM
I use Vegas. But I don't think you can bet around the time stretch factor in any package on a series of images. And in general I always use image sequences as an intermediate step. I always work in an avi of some sort. I don't see anything wrong with your workflow other than trying to use sequences in the timeline. I am not sure that is a great idea.

I think you can just use a standard avi compression codec such as Cinepack Codec by Radius. Or any one of your choice. You can play with the quality settings. Or you can use any compression of QT as mentioned. It does not have to be DV. Just compressed avi.

Load up the sequence into Premier and render it out with one of the supplied codecs.