Thank for your reply!
So I don't play with MXF and continue as original m2t files?

I've made a reasoning like that :
my HDV is 25 Mbps,
if I render as 35 Mbps and if I make BD's as 25 Mbps,
the image quality, sharpness etc. will be better??
But I undersand that this is not the case!

That's right, you simply cannot in a technical sense make it any better than what the camera recorded. You could transcode it to uncompressed and you gain nothing, you don't loose anything either but you'd chew up a lot of disks doing that so why do it.
Anything other than uncompressed must make it worse because they're lossy codecs. You might transcode because it's easier to edit using a different codec and the loss might be very, very small but it will absolutely not be better than the camera original.

Bob.

In a "technical sense," you can't make it any better, but if you transcode to a codec with better color sampling, such as from 4:2:0 to 4:2:2, there can be chroma smoothing where the codec adds "interpolated color samples."

This can generate additional perceived quality, making the output look richer (how much depends on the codec implementation), while you stay in 4:2:2.

Also, by transcoding to a less lossy codec you can reduce losses in grading and compositing (depending on what you do).

So it sometimes makes sense to transcode, it's just another tool to be pulled out when needed.

What Bob says about a transcode not improving on the original format is true, but I transcode anyway because I can't see this loss and the transcoded footage is much easier to deal with.

Not only that, but color corrections look better. MXF and Cineform may round up or down the color numbers when they first transcode, but subsequent color corrections are starting from a more exact place and you can see the difference. Look on the gradual color gradations in the sky to see what I mean.

Then there's the fact that both MXF and Cineform smart-render. You may have an extra generation up front, but by the time you go to your master you'll likely have recovered that generation because of the smart-rendering capability.

Both MXF and Cineform have uncompressed audio. Every time you rerender into some other compressed codec (like m2t mpeg2) you decompress and recompress the audio, even if you smart-render the video (as is the case with m2t mpeg2).

In the case of Cineform, color correction is limited to 8 bits in Vegas. Cineform's First Light color corrects at 10 bit precision with highly optimized and very fast code. What this means is that if you edit in Vegas but color correct with First Light you are getting 10 bit color correction which Vegas then truncates to 8 bits. This gives you much more precise color correction grading: kind of like what the 32 bit modes in Vegas do but without the huge render times and constant crashing.

..but if you transcode to a codec with better color sampling, such as from 4:2:0 to 4:2:2, there can be chroma smoothing where the codec adds "interpolated color samples.

I agree with the overall idea but I think the process goes more like this:

8-bit 4:2:0 is decoded. Interpolated color samples (actually probably 4:4:4 while in RAM) are generated at this point.
8-bit 4:2:2 is then encoded and (some of) those previously interpolated color samples are recorded to the new file.

Regardless, the image data is a bit more ample. It shouldn't look richer yet but the image ought to hold up to manipulation better over a couple of generations. Still, if generation loss is a concern then something like Cineform or maybe DNxHD is a better choice.

Rob Mack

Rob,

You just described "transcoding to a codec with better color sampling."

Unless you are working entirely in RAM (where 4:4:4 is used whenever possible, usually on a frame basis) and stay in RAM, transcoding means reading one file, processing the frames and pumping out a new file in the new format.

That is the process.

The transcoded media can look better immediately, if you're viewing it on a format-appropriate monitor with 1:1 pixels and proper output from your computer in the actual video format (as opposed to say looking at the Vegas preview window on a computer screen).

See this for what I think is the cheapest way to do it right.

Keep in mind that a few years ago, all the Windows NLEs used VFW and hence RGB color. Now Vegas is pretty much the only straggler left still using this archaic approach. This means that every time you render the video you are going through another generation of YUV to RGB color conversions. Using a better colorspace format may not look any different on the first generation, but it looks a heck of a lot better after multiple generations of this YUV to RGB to YUV to RGB, especially if there is color correction involved.

For example, I have no problem whatsoever putting m2t mpeg 2 video on a Vegas timeline. I absolutely never render to it unless it is for a specific use like authoring a Blu-ray disc or playing back on a Playstation 3. For anything that involves rerenders I'll use Cineform preferably or MXF as my second choice.

" the transcoded footage is much easier to deal with."

Transcoding from one mpeg-2 codec to another, higher bitrate one is unlikely to make it easier to deal with.

Indeed there are processes that will make footage more appealing to the eye, that's why I said in a "technical sense". The thing is they should be done once only and this gets overlooked. Apply multiple generations of chroma smoothing and things can go downhill quite badly.


"Every time you rerender into some other compressed codec (like m2t mpeg2) you decompress and recompress the audio, even if you smart-render the video (as is the case with m2t mpeg2)."

Going from HDV's mpeg2 to uncompressed audio in MXF does not remove the compression artifacts. There's no need to go through the process, you gain nothing. In the case of the audio you neither gain or loose anything, you just do more work for zero yield. Going straight from what is recorded on the tape to your final format is as good as it can get, ever.

Bob.