Editing raw footage

[Michael Chen]

Hi all,

I know adding effects to a clip such as transitions, colour correction and so on will cause minor degrades to the footages.

But what if I am only cutting and pasting the footages?

I would just like to do some cutting and pasting in my computer and output it to another computer for colour correction.

Will there be any quality losses?

[Edward Troxel]

Cuts only should result in ZERO loss of quality. Rendering will result in the file being COPIED - not recompressed.

[Michael Chen]

Well, I guess thats what I needed.
Thanks Edward!

[Alex Knappenberger]

Yup.

Also, Sonic Foundry's DV codec (used in vegas) is found to be much better then Microsoft's (used in premiere and the others). In a test, SF's Codec could re-render (not sure of what word to use, possibly recompress) the footage upto a infinite amount of times without seeing any degrading video quality, in the test they redid it 50 times. However, after about only 15 recompressions, you could see significant degration in the Microsoft encoded footage.

[Jeff Donald]

Do you have a link to that test?

The moral of the story is do not render your scene 15 times. Be aware that rendering your scene causes a slight lose of quality each time. The quality lose is cumulative. You may not notice the first few times you render, but the artifacts are sneaking in. Keep your changes to a minimum and render as few a s times as possible.

[Alex Knappenberger]

Here it is:

http://www.digitalvideoediting.com/2...vegasvid3.htm#

Directly to the sample picture with comparisons:

http://images.digitalmedianet.com/20...vid3/codec.jpg

[Michael Chen]

ok how bout adding sound effects / voice overs?
It doesnt add any effect on the video itself, just on the sound part.

Will the quality be worse off compared to if i just do cuts and trims?

[Ed Fiebke]

Up until now, I was operating under the assumption that there wasn't degradation when scenes are rendered and re-rendered again after more effects, etc. However, after viewing the sample pictures offered by Alex and reading the comments to this thread, I've now learned otherwise.

Slightly confused, I seek clarification:

1) Does the degradation process also include copying the video signal from a DV (miniDV) tape to the computer via IEEE 1394?

2) Are editing cuts and splicing made during the editing process considered some form of rendering and therefore subject to degradation too? (To be honest, I don't fully understand how the process works whenever I "shorten" or cut out parts of a scene on a particular video footage. I just take for granted that it does works and marrily splice and edit away.)

Now that I understand that degradation of the video occurs with each repeated rendering, I ask the following questions:

1) What type of editing/effect processes causes the most severe degradation during repeated rendering?

2) What DV editing techniques to you used to help keep the degradation process to a minimum during repeated rendering of edited scenes?

Great topic, and I appreciate the knowledge shared here!

Ted

[Alex Gingell]

Capturing to computer via firewire causes no quality degradation...it is merely copying the already compressed (by the camera - dv is 5:1 compressed) files across.

AFAIK editing by purely slicing and dicing your .avi's is not going to change a thing because when it renders out it should just be copying portions from all the .avi's to your final .avi.

Where you will get quality losses is if you re-render your video in an effects package like after effects or 3dsmax for purposes of compositing or adding vfx (as I learned harshly myself).

My question is, if you do all the effects work with uncompressed files (i.e. you import your starting dv .avi's and then all renders are to uncompressed .avi's) can you get away with no loss in quality UNTIL you have to render out in the final step back to compressed dv?

Assuming you use a good codec like mainconcept dv or sonic foundry that last compression step should be your only loss in quality, and hopefully not by much!

After viewing this:

http://images.digitalmedianet.com/20...vid3/codec.jpg

I can safely say that $50 for mainconcept's dv codec is looking mighty tempting. Does mainconcept's compare to sonic foundry's ? Is sonic foundry's available outside vegas ?

Alex

[Rob Lohman]

I'm not really keen on this comparison between codecs, why?
Because of the following reasons:

1. it was done 1,5 years ago
2. it is known that certain Microsoft DV codecs were bad and others were a lot better, they don't specifiy which MS DV codec they used
3. they don't talk AT ALL about how they run the tests with what applications and with which source material.
4. there are no fullscreen samples available to check out yourself in an UNCOMPRESSED or LOSSLESS compression format. Who says their resizing routines and jpeg saving is good?

Now I'm not saying that the Vegas codec isn't better than the
Microsoft one or Mainconcept... But what I like to see is the
following:

1. a test that has been done NOW (not a couple of years ago)
2. a test that clearly explains how everything was tested and with which source material and such (ideally with multiple images)
3. a test with more than one codec and different versions (Microsoft DV codecs on Windows 98, 2000 and XP for example. Vegas codec 3 and 4 etc.)
4. output images that are not compressed and full resolution
5. unbiased (I am not saying that the tests above are biased, sponsered or whatever)

I am thinking of perhaps doing such a test myself for the Watchdog,
but currently time and a not clean PC with multiple OS-es is
prohibiting me.

[Jeff Donald]

I agree with Rob. I'm not a big MS fan, being a Mac user, but a lot changes in a year and a half. I would think the gap has narrowed quite a bit from what I've seen of the the MS codec. The VV codec is acknowledged as one of the better Codecs in the industry. Most people would never render a scene 50 times and with careful media management, should be able to limit the number of renders to well under 10.

[Robert Bobson]

Just to clarify, if you're editing uncompressed DV and exporting back to tape, does this degradation apply?

BB

[Jeff Donald]

DV is not uncompressed. It is compressed 5:1 per the DV specification. When you watch the image, the codec decompresses it for viewing. This applies to either analog viewing via S-Video or RCA. There are very few FireWire display devices at this time.

If you capture your DV footage via analog, the capture card could capture at more or less compression depending on the codec used for capture and various settings. It could be captured as uncompressed files also (about 30mb per second or 108 gigs per hour for SD video).

DV could be captured (transferred might be a more accurate term) then rendered in an uncompressed format. Once uncompressed it could be edited uncompressed and all effects, graphics etc. rendered uncompressed. This would yield high quality files with no compression artifacts.

Is this feasible? For most people no. Standard IDE or FireWire drives are not fast enough. SCSI RAID 0 drives are the standard for this type of work. Uncompressed capture cards are expensive.

Once you have your finished project what are you going to do with it? If you go back to DV you're compressing it again (5:1 per the spec). You need a high end VTR (D-1, D-5, Digi Beta) to record your signal and again big bucks.

[Don Donatello]

nobody renders 50x ... i do a ONE render test

if you want to see the difference in how codec's handle color
load in NTSC smpte color bars
take a digital readout of all the bars ( for this post just RGB)
these are what they should read for SMPTE
RED r180 g16 b16
GREEN r16 G 180 b16
BLUE r16 g16 b180

render out the bars
if bars are media generated ( digital - like in Vegas) ) you just render them out
if bars were from a camera you'll need to force a re-render ..

results of ONE render
MS= microsoft DV codec
SOFO = vegas codec

RED 180 16 16
MS 173 29 28
sofo 180 16 16

Green 16 180 16
MS 27 171 27
sofo 16 180 15

Blue 16 16 180
MS 29 30 174
sofo 16 16 180


now look at all 3 bars on vectorscope and waveform
compare MS, SOFO to the digital generated bars

NOTICE that on the waveform every MS color has loss luminance.
on the vectorscope every MS color loss satuation. SOFO on waveform hits all bars on the mark , vectroscope all the color's hit their mark ...

MS also made a mess of your blacks
the 3 pluge bars ( 3.5 , 7.5 , 11.5 ire )
digital RGB readouts
RGB for 11.5 is 41 41 41
MS 51 51 51 , SOFO 41 41 41

RGB for 7.5 is 16 16 16
MS 30 30 30 , sofo 16 16 16

RGB for 3.5 is 7 7 7
MS 22 22 22 , sofo 7 7 7

100% white RGB 235 235 235
MS 218 218 218 , SOFO 235 235 235

also the MS has artifacts from the cyan bar running into the small magenta below it ..and the magenta bar has artifacts running into the small cyan below it
i see no problems with SOFO ..


running V4 , microsoft directX 9 ... using V generated smpte NTSC color bars
for V users - use the cookie cutter FX ..it has a eye dropper to sample colors for RGB read out

[Simon Orange]

Don,

Very interesting. The MS codec seems terribly lacking !!

Although I do not in any way doubt your results, I would like a little more detail before drawing any conclusions....Rob has outlined these earlier. You have not mention which version of the codecs you were using...what OS..what you were rendering in for example. I think it would be worth doing the tests with the Mainconcept codec also (anyone interested in trying) or even the canupus one - or is that a variation on the MC one ?

With regard to the 50x render tests mentioned above, I am suprised that the images using the VV codec looked so good after 50 recompresses. I would have imagined that the picture would have degraded appreciably even with a great codec (we are talking 5:1 compression here). This leads me to suspect that there 'might' be something else going on in vegas (?).

Just to add my 2pence to the whole discussion. Premiere, by default uses the MS codec. This means that capturing in premiere uses the MS codec, but this isn't really an issue as in this instance it is just acting a a wrapper to get the data from the camera to the .avi file. There is theoretically no quality loss on this initial transfer. If you then go on to edit only applying cuts...not dissolves or anything that needs rendering....then lay it back out to tape there is NO quality loss.

The issue is with FX and transitions that need rendering. It is more than feasable to do bits of FX work uncompressed using either uncompressed TIFs or AVI, then you will get the best quality possible. It will be near to impossible to play uncompressed video media on your average PC, but you can either preview to RAM or make a lo-quality proxy. If you can possibly manage it keep all your data at the highest 'practicable' quality for as long as possible and only compress once - at the end.

Unfortunately at some stage 'if' you want to lay back out to DV tape you will need to render down using a DV codec. This is where the of course where the codec quality issue comes in.

If you are using something other than Premiere and are not happy with the MS codec (test results pending) then buy the mainconcept codec (ditto). In premiere, however, the MS codec is quite deeply embedded. There are tweaks you can do to premiere to force it to use the mainconcept codec - this is something that some of you might want to (do a search on the internet...and the adobe site and abcdv.com..look for qcap.dll)

PS all of the above applies to PCs only...macs are a whole different ball game !

simon