Vegas 10 - to transcode or not?

[Geoffrey Chandler]

I found this article How to Transcode .mov Files using Matrox Codecs | Custom Made Computers

-----------------------------
"There are different codecs for different purposes. There are capture codecs (H.264, AVCHD, etc.). They encode fast but decode slow. Thus, they’re excellent for capturing but terrible for playback. There are editing codecs (DNxHD, Matrox, Cineform) which create large files but which don’t lose much information during the color grading / editing processes. Then there are delivery codecs which compress efficiently while maintaining visual quality. These files are high quality and small.

Professionals will transcode for each step of the process, unless they shoot something at four o’ clock and they need to roll out a finished product by five o’ clock. But to maintain optimal quality, transcoding for each step is necessary. It is often possible to edit .mov files directly, but if you do much color grading and editing you’ll end up with a far inferior product than if you’d transcoded to an appropriate intermediate format."
-----------
So, even though Vegas 10 seems to handle H.264 files out of DSLRs better than before, would it be advisable to transcode before editing anyway, for the highest quality product??

Thanks,
Geoff

[James Millward]

I would like to know this too...

I dont quite understand why editing in the native h.264 would not yield the best results. After all it is the original file.

Surely anything we do it via transcoding can only result in degredation on some level??

James

[Rainer Listing]

If I sell codecs, what am I going to tell you? Instead, think: transcoding from one lossy codec to another lossy codec, what's going to happen? The key is in the "if you do much color grading and editing". Converting to a format with say a 4:2:2 colorspace will give you a bit more latitude for color correction, and if you are mixing footage from different sources it will create fewer problems if everything is the same. As for professionals transcoding, if they are on FCP or Avid, which most of them are, they don't have any choice. If you are happy with your footage as shot, and your system can handle it, you retain maximum quality if you don't transcode your input.

[Leslie Wand]

interesting topic, and one that's going to be so ever more ;-)

i rarely transcode, but then again i'm usually editing m2t, and the usual suspects. when i do have avchd (which edits fine on my pc) i normally leave it raw, only when i intend doing anything 'serious' with it (eg. cc'ing, fx, etc.,) do i transcode, and then to mxf.

haven't had much experience with dslr, but if it played happily on the tl, again, unless i was going to manipulate it, i'd leave it alone.....

the following is MY opinion after 30+ in tv and video production - the less i manipulate / transcode material the better. and, for a great deal of basic manipulation (very basic contrast / brightness, etc.,) it really isn't worth my while transcoding. whatever loss might occur with the original ISN'T going to make that noticeable a difference by the time it gets to dvd.

since i don't (as yet) deliver hd on b-ray (just via net), again, i don't really see the need to transcode. there's a great many perfectionists out there who feel ANY loss is a major 'problem' - well, if my audience was made up of pros examining my work for artifacts, (very) slight softening (relatively to original footage), then i might worry, but as it is, the vast majority of my audience is made up of people used to watching youtube quality masquerading as entertainment, poor satellite, and mobile phone on the news. frankly, whatever i produce looks great in comparison - and is only bettered (in their opinion) by studio releases on dvd.....

[Edmond Engler]

Just for a little bit of fun I made two virtually identical videos.

I didn't edit these clips at all. They're straight out of the camera. All I did was string them together in Sony Vegas Pro 10.

The only difference between these two videos is that one of them was made with matrox .avi files. In other words, I transcoded the .mov files to matrox .avi files and then I rendered to mp4 in Vegas Pro 10. I used the highest bitrate (300 MB/sec) that the matrox codecs would allow, and my render was 8,000,000 bps.

The other video was made with the original .mov files straight out of the camera. Again, I didn't edit the clips at all . . . I just strung them together.

So, to sum up, I made two identical videos. I didn't edit the clips at all. I didn't color grade or make any changes to the clips.

Can you tell which is which?

Transcoded or Not? on Vimeo

Transcoded or Not? #2 on Vimeo

[Leslie Wand]

ok, i'm willing to be proved wrong - my ego can take it ;-)

the second clip is sharper, and since you say you didn't do anything to the original footage i'll say the second is camera original and the first transcoded.

if it's the other way around, then there's some sharpening going on it the transcoding which might end up giving the final edit a 'brittle' look.....

ok, what do i know.....

[Edmond Engler]

The first clip (Transcoded or Not #2) is the transcoded clip. I find that the sharpness is cut down just a bit, but the scenes are also less contrasty (the shadows aren't so black) and I like that.

I still haven't decided whether to continue transcoding to this intermediate format or to edit the .mov files directly. I don't think there's any noticeable loss of detail (with the matrox version) and I do believe that with a lot of color grading, etc. that using an appropriate intermediate file is preferable.

But hey, I'm no expert. Not by any means.

[Leslie Wand]

frankly i think there's much too much attention paid to transcoding and not enough to that which is being transcoded ;-)

well exposed / shot material that doesn't require heavy post should be edited as is, be it hdv or avchd (if pc can handle it!). transcoding is, in my opinion, a rather time consuming and pointless exercise if the workflow is pretty much camera to dvd.

if i do have to transcode (rarely), i use mxf - it's free, works well in vegas, and can handle the sort of cc'ing and stuff i do quite adequately.

however, IF you're looking at heavy duty chroma keying, complex compositing, etc., you're better off with 4:2:2. YMMV

[Phil Lee]

Hi

This is something I've wondered about for a while, transcoding to 4.2.2 for clips on the timeline for better eventual output

Say this is my workflow:

1) AVCHD from camcorder, not transcoded to anything

2) Various edits/transitions/colour correction applied on the timeline

3) Render out edited footage as Sony YUV, or go direct to Blu-ray/AVCHD from Sony Vegas.

In the case of the uncompressed Sony YUV I go on to make HD or SD footage using an alternative encoder.

Now my understanding is at point 3, when the movie is being rendered out and having colour correction/effects applied the video is uncompressed from AVCHD to at least to 4.2.2 or even 4.4.4? The effects are applied and then would be wrapped up in Sony YUV 4.2.2. So all compositing/chroma keying will have happened in at least 4.2.2 colour space anyway, regardless of the original source colour space.

If the above is true, what I fail to understand is how transcoding AVCHD to some form of 4.2.2 format and placing that on the timeline, makes any difference to the final output to make it better?

I can see where having the video in 4.2.2 would be better if that output then forms the input to something else for further work, so in 3 above you wouldn't output as AVCHD again, but wouldn't see the point of transcoding at step one first, just do it at the output stage, so that kills two birds with one stone.

Just wondered what people thought.

Regards

Phil

[Craig Longman]

First off, I think there is a case for transcoding, and a case for not.

The article is essentially correct. On capture devices, the codes are designed to encode very fast, but they invariably are expensive to decode. For editing, it helps to have one easy on the decoding, as lossless as possible (big files) and keyframing handled in such a way as going backwards and forwards frame by frame is easy for the codec (not easy for many of them). For playback, it doesn't matter how long (technically at least) it takes to encode, but it should be as fast as possible to decode and as small as possible.

The case for transcoding, is IF you need to transfer your files around between programs, or for some reason need to render in stages. Then, a codec designed for transcoding is far, far superior.

The case for not, is if you have a computer powerful enough to handle the source codec, if you bring your footage in, cut/edit/splice it up, apply one or two filters, then immediately render it right out. In this case, you'd gain nothing by transcoding; you would, in fact, lose some (perceivable or not) quality.

And for Phil, the question is more about what quality it came in at. Inside Vegas, the plug-ins have all 3 colours and an alpha channel for each and every pixel, essentially 4:4:4(:4) quality. If you only have a DV (4:1:1) quality source, then it's still only going to be calculating the RGB values based on 4 luma values and one set of chroma values. That's true whether you transcoded to 4:4:4 or not. Now, mind you, the one exception would be if the codec provided some sort of actual chroma sampling up-conversion (as opposed to just a chroma blur), in that case it might help.
But even still, I think the same rules apply. If you do stuff in Vegas (or any editor) then render in multiple passes or transfer between apps requiring rendering, then an intermediate (and/or increased chroma sampling) will help. If you just bring it in once, do you bit, then render for delivery, it really doesn't matter.

[Phil Lee]

Hi

Thanks for your reply.

So apart from performance benefits of a less compressed easier to work with codec, there is no quality improvement to be had moving from AVCHD->Timeline->Final Output, which sees one decompression, one compression as opposed to going AVCHD->Transcode to something easier to edit, bigger colour space->Timeline->Final Output, this then means you're video has gone through 2 decompressions and two recompressions, so you will end up with worse quality in theory, in practice if you are using a good intermediate codec visually it might not be noticeable, in the case of the posted comparison it was spotted which was the transcode one.

This confirms what logic was telling me, however there does seem to be general belief that transcoding to something like CineForm Neoscene with a bigger colour space, automatically gives you a better final output dropping that onto the timeline rather than working with the native AVCHD.

Regards

Phil

[Leslie Wand]

what you record is what you record - here's no software that will add what isn't there.

you might tweak whatever there is, and (with transcoding) preserve whatever's there to some degree while heavily manipulating the data - but at the end of the day, it's down to the digital quality of your original footage.

[Craig Longman]

Originally Posted by Phil Lee

[...snip...]
transcoding to something like CineForm Neoscene with a bigger colour space, automatically gives you a better final output dropping that onto the timeline rather than working with the native AVCHD
[...snip...]

I do only DV editing, and it's all 4:1:1 color sampling (1 chroma sampling for 4 pixels wide) and it really is hard to see that. 4:2:0 (which I think is what AVCHD is?) is even better being 2x2 pixels.
The time 4:1:1 really sucks is when you're trying to chroma key something out. I tried out the NeoScene HD product (the $500 one, is the only one that handles SD footage in/out) and it made a huge difference, clearly doing some fine interpolating while up-sampling to 4:4:4. I decided it was too expensive for me just for chroma keying, especially when a separate mask and a bit of chroma blur does a pretty decent job.

The reason they do encode at 4:1:1, 4:2:2 or 4:2:0 is it is _very_ (sometimes very, very) hard to visually tell the difference. Our eyes are just not that adept at picking up colour differences, not nearly as much as picking up intensity (or luma) differences.

CineForm is a good product, and has numerous advantages other than just being a great intermediate codec. The First Light part of the HD product is great, and if you do need to export to AE or something else and back, there's nothing better.

But to simply buy it to be able to say your source file is 4:4:4 I don't think makes any sense. As Leslie said, it's more about what goes in that anything else. (although in specific technical cases like chroma-keying, software _can_ "add"/interpolate what wasn't there pretty well =)...

[Adam Stanislav]

Originally Posted by Craig Longman

Our eyes are just not that adept at picking up colour differences

Speak for yourself! :-)

I think the real reason why they came up with reduced chroma sampling was that tapes were expensive and TV bandwidth was limited. They figured that the audience could not tell the difference not because their eyes were not good at differentiating colors but because they were not there, so they had nothing to compare it with. They actually did that when they started with color TV, when the colors looked terrible anyway, so no one could tell 4:4:4 from 4:1:1 (or, actually, their analog counterparts). Then when early digital video came, digital tapes were still expensive and the result was shown on analog TV sets, which by then were better but people were still sitting so far from them they could not see the difference. And most home TV sets were not properly calibrated anyway.

But let's not kid ourselves that we cannot tell the difference between 4:4:4 and 4:1:1 with present day high definition digital video. Because we can. Perhaps not everyone, but anyone who works with colors certainly can.

We color grade our images not to get a precise record of reality but because we want to create our own reality. So, it is not whether we get the color of each pixel to show exactly what we shot but whether the image looks the way we want it to look. And when creating a look, we are working with the entire image and not with individual pixels.

And that, to me, is the true and honest reason why 4:2:2 and 4:2:0 and 4:1:1 are still used today and not some deficiencies of human vision.