AVCHD is lossy?

E

EmeraldICE

Member
From what I understand, editing AVCHD is lossy. So does that mean I should be transcoding to a different format before editing?
I've use neoscene to remove pulldown on 24p clips but should I be using it for all my footage?
Is the neoscene export lossless? Or is that lossy too?
 
Every codec is lossy, unless it specifically specifies "lossless". But the losses only come into play when you recompress. So if you put raw AVCHD files on your timeline, they're no more lossy during editing than uncompressed files would be.

In the editing process, uncompressing compressed footage doesn't introduce any manner of loss. It's only the recompressing that matters. So if you do straight edits, or if you don't render down portions, you'll never be introducing loss to your images.

When you render, the codec you render to will introduce its own level of loss into the process. Cineform is less lossy than AVCHD, which can easily be seen in the size of the files it generates, Cineform files can be 10x as large as AVCHD files. That additional file size helps it to preserve more detail during recompression.
 
Barry_Green said:
It's only the recompressing that matters. So if you do straight edits, or if you don't render down portions, you'll never be introducing loss to your images.
Click to expand...

Barry, your expertise is always appreciated. I don't understand two things:

Why would rendering straight edits not result in quality loss? How can you exit an NLE with a video file that has non-rendered portions?
 
If you're doing cuts-only editing, nothing gets re-rendered. If you export to uncompressed, nothing has been compressed in a lossy codec.

These aren't necessarily typical workflows, but they do help to illustrate the point.

If you put in a cross-dissolve or something like that, then there will be rendering done on that, so it depends on what you're rendering to. If you're rendering to a lossy codec, loss will happen. If you're rendering to a lossless codec, no loss will be incurred.
 
Ahhh. Thank you for explaining it Barry. That makes sense.
So basically I'm just making things more complicated than they really are. Typical.
My machine handles AVCHD just fine, so I guess there's no need for me to transcode it to something else.
 
The only benefit in transcoding is that you'll get snappier, faster editing performance.

The drawback is that it takes some time to transcode, and the transcoded files will be bigger (perhaps 10x to 20x bigger). So it's up to you to decide. For some people the timeline performance gained is well worth the file size tradeoff... for others, the immediacy of doing a same-day-edit prevents them from taking the time to do a transcode first. So it's up to you.
 
Barry_Green said:
... If you export to uncompressed, nothing has been compressed in a lossy codec...
Click to expand...

I've operated based on Stu Maschwitz's dictum, "No mastering in the NLE." He writes (p. 246, The DV Rebel's Guide) about lossless video codecs, "...the truth is those codecs aren't actually lossless. They're usually 4:2:2...or compromised in some other way."

Is this information outdated? What are the "other compromises?" Are you saying that it is possible to render losslessly while transcoding (in addition to transcoding losslessly prior to editing)?
 
There's a lot of nonsense thrown around with the word "lossless". Some people think Redcode or Cineform or Prores are "lossless", and they're not. Then people cling to the notion of "virtually lossless", and by that they mean: if you can't see the difference, then it's lossless enough, right?

True lossless, like a .bmp or .tga file, involves the ability to restore the image identically to what it was before compression. And lossless compression does exist; heck, everything that .ZIP or .RAR or StuffIt does is lossless.

But they're not all that commonly used in video editing applications.

I don't know what other compromises he's talking about. Typically a lossless codec gets only maybe 20% to 40% space savings, or perhaps less, so that's certainly a compromise; shuffling around a gigabit per second of data is no small task for the vast majority of computers. So maybe that's what he means by "some other way"? I don't know, you'd have to ask him.
 
jordandov said:
I've operated based on Stu Maschwitz's dictum, "No mastering in the NLE." He writes (p. 246, The DV Rebel's Guide) about lossless video codecs, "...the truth is those codecs aren't actually lossless. They're usually 4:2:2...or compromised in some other way."

Is this information outdated? What are the "other compromises?" Are you saying that it is possible to render losslessly while transcoding (in addition to transcoding losslessly prior to editing)?
Click to expand...

He's probably referring to colorspace loss during subsampling (or even interpolation to a larger chroma resolution... it's still TECHNICALLY a loss).

Thing is, unless you're shooting with a higher end rig, or are scanning in film to a 4:4:4 DI, you're most likely working in 8-bit, 4:2:0 (HDV, AVCHD, PAL DV) or 4:1:1 (NTSC DV). DVCProHD uses 4:2:2, so you're starting higher; you'd only lose if you went down to 4:2:0 or 4:1:1.

Of course, bringing in bit depths, that's another potential for loss, and in my mind, a far greater loss. The chroma resolution does not give you as grand an advantage as bit depth does (the exception being chroma keying), which he could have also implied when transcoding, though again, most sub-$5-6000 cameras shoot 8-bit anyway.

Finally, DVDs and Blu-rays are always 8-bit 4:2:0 anyway, so if you're maintaining the same colors throughout your workflow, you shouldn't worry about it.
 
most sub-$5-6000 cameras shoot 8-bit anyway
Click to expand...
There's only one 10-bit camera on the market under $10,000, and that's the HPX300, and even then only when shooting to AVC-Intra. So unless you're shooting AVC-Intra on an HPX300, you're starting and ending with 8-bit.
 
Barry_Green said:
There's only one 10-bit camera on the market under $10,000, and that's the HPX300, and even then only when shooting to AVC-Intra. So unless you're shooting AVC-Intra on an HPX300, you're starting and ending with 8-bit.
Click to expand...
Ahh, I was under the assumption that the HVX was 10-bit in addition to being 4:2:2, my mistake!
 
I would think because of the headroom you have. You'd probably end up degrading the picture much more quickly while editing when working in an 8bit environment as opposed to a 16bit environment. I'm just guessing here.
 
Ian-T said:
I would think because of the headroom you have. You'd probably end up degrading the picture much more quickly while editing when working in an 8bit environment as opposed to a 16bit environment. I'm just guessing here.
Click to expand...

That's only true to the extent of color grading. You're not degrading in any other regard usually. Noise removal also becomes an issue since you lose noise/film grain that acts as a dither on your 8-bit video, hence posterization on a lot of noise filters.
 
You must log in or register to reply here.
Back
Top