DV: Losless conversion?

[dynamique]

I wonder if the quality is lowered if I import a dv clip exported from one dv NLE to another dv NLE and export it again. Does it matter if the underlying codecs are different? Let's say I do it like five times.

[redwudz]

This was discussed some time ago on our site, though I don't remember the thread title. There should be almost no generational lose with 5 exports and imports, if nothing is altered in the file. This could change if different DV codecs are used with each process. But in normal practice, exports and imports shouldn't cause losses.

I'm not familiar with how each one of the available DV codecs works, but if they do color space changes or similar, I could foresee problems. I'll leave the explanation of those technicalities to one of our members more knowledgeable about the codecs and how they work.

[dynamique]

Actually my conversions would include some changes into the video material.

[redwudz]

Since it varies with the programs and codecs used, you might make up a short clip and run it through several import/export cycles and see if there is quality loss. Then you could determine if it's the codecs used or the encoding.

EDIT: There's a few threads concerning DV and colorspace in our DV / HDV forum. Try a search there for 'colorspace'.

[Cornucopia]

5 times (with modifications each time) is pushing it (no matter which DV codec brand). Tests at adamwilt.com have shown that ALL codecs will be showing quite noticeable degradation after 9 passes. 5 is more than 1/2way there.

What are you doing that requires multiple NLEs, that can't be done with one or 2 passes? (Track layer/3D fx limits?)

If you absolutely have to run so many passes, make your 1st encode be saved as Lossless (HuffYUV, Lagarith, etc) and keep it that way until the encode out of the last pass.
Yes, it's going to take LOTS more available space, but that's what one does under such unusual circumstances.

Scott

[dynamique]

Thanks.
redwudz, the thread you pasted doesn't hold the answers, just the search function.
Well, it's more a kind of knowledge research I' doing. But I have to crop a finished movie, and I want to keep as high quality as possible. I have already re-mastered it like two or three times.

[Cornucopia]

"Cropping" should only be 1 pass through 1 NLE.

Scott

[dynamique]

With "1 pass" you mean one conversion? I am kind of a newbie to the exact meanings of conversion terms.
I know a compression can be one or two-pass but I thought that meant how many times the compressor passes the material to calculate the new pixels using the algoritm (where more passes are positive to the material so to say)

/Nobody tought me these things in primary school

[celtic_druid]

Generations might be a better term.
DV capture is lossless, but the codec itself is not, so if you re-compress, then there is quality loss. Basically it is like resaving a jpeg multiple times.

Also different NLE's have different smart rendering capabilities, but in theory where the video hasn't changed (since DV is all keyframes) the video should not be compressed and if you are working on different sections in each app, then no section would be compressed more than once. In the past I have noticed that with some apps if the FourCC (DV can use more than 1) isn't right then it would recompress all video.