Transitions (Disolves) are DARK

crombej wrote on 8/23/2001, 1:14 AM

Hi,

Using a Dazzle DVC II card, capturing video (Svideo) at 720x480 Mpeg2, 8mbps. 800mhz P3, 512mb ram.

Ok, captured video (simple one is only 75mb for 2 minute video) looks fine. Drag onto VideoFactory 2 and all is good (although it won't show preview in real-time but that's to be expected I guess).

Now, I simply go to a point, say 5 seconds into the video. Do a SPLIT. Now drag the right clip over the left clip for say, 15 frames and I have a cross-disolve (I have that button/switch turned on).

Here's where the trouble starts. When I now click on transition area, my video preview window is usually DARK. It's not completely black but many of the frames (but not all) look like they're only about 50% of the brightness they used to be. When I render the clip, the transitions always flicker and have a short black section in them.

However! If I do a Make Movie and create an AVI file from the original MPG, then repeat the above process, the transition is as smooth as silk - all the same brightness level and perfect disolve into the next piece.

I've tried right-clicking on the MPG clip in the media bin and changing the field order from the current value (lower first) to upper first or progressive but that doesn't seem to do it.

What's the deal? Do I have to render my 2 hours of video into AVI format first? Please no! It would be like 100gb at least.

Any help would be greatly appreciated!

I can post a MPG file that shows the problem if that would help?

Thanks very much in advance,
JC

Comments

Chienworks wrote on 8/23/2001, 8:30 AM

I'm sure the problem is that your source clips are MPEG format. MPEG is
highly compressed, and does not store the complete image of each
frame. Most of the frames have to be reconstructed from occasional key
frames and bits of information on how the current image differs from
that frame. When you have a transition, you are combining two MPEG
streams together and doubling the amount of processing that must be
done. This is a lot of work for the processor to do and most preview
systems don't bother trying to do this.

Capturing your video in AVI format to begin with would certainly be
preferable as it would avoid your preview problem and potentially give
you much better quality in the long run. However, as you say, that would
be a huge file. Even using the YUV9 encoding, you would probably need
0.5GB per minute. You could however capture the video clips in pieces
and then render them to DV format. This is still an uncompressed AVI
format which only requires 0.225GB per minute.

If you simply want to see your transitions without rendering the entire
project, you can select just the portion of the timeline containing the
transition, then check "Render loop region only" to render just that
transition. This should only take a few seconds, dending on how long
the transition is.

crombej wrote on 8/23/2001, 11:04 AM

Hi,

Thanks for the help. A couple points of clarification.

When I get those dark / flicker parts in my transitions, nothing can fix them. I do pre-render - no help. I do a full make movie - still no help. They're really in there! I can post a fairly small clip which shows this on my web site if anyone wants to see it?

Next, you said I could render it as "DV" which is lossless at .225gb/min. That brings up two questions:

1) If DV is lossless, then how come Video Factory has DVD "Good", "Medium" and "Better"? Why are there different compression rates if it's lossless? Help!

2) How do I render as a lossless "DV" format with MakeMovie?

Still the big question is why am I getting this crazy flicker/dark sections on my transitions?

Thanks again for any help!
John

discdude wrote on 8/23/2001, 11:16 AM

DV is not lossless.

SonyEPM wrote on 8/23/2001, 11:47 AM

I think this is an MPEG-to-MPEG transcoding problem.

Try rendering a short file to avi using the NTSC DV template- are the dark portions still there?

crombej wrote on 8/27/2001, 12:01 PM

Hi all,

I think I have an idea what's causing this problem!

It looks like it only happens when I split BETWEEN frames! If I use ALT-RightArrow (or leftarrow) to make sure I'm on an even frame boundary, it doesn't happen. However, if I'm zoomed in pretty close and just click, or just use left-arrow or right arrow, I'm usually NOT on a clean frame boundary.

This is when I see the problem.

This only occurs with my original video which is "Upper Field First" field order.

Seems like others would have run into this before?

Well, any help would be appreciated, but it looks like I can work around this by alt-right-click first.

Can you force Video Factory 2 to ONLY allow cursoring to frame boundaries? Snap doesn't seem to be enough. What good is right-arrow / left-arrow anyway? A "pixel" movement doesn't mean anything???

Thanks!
John

wvg wrote on 8/27/2001, 12:14 PM

Are you sure you are not introducing blank frames?

I've notice a similar problem, whereas if I split the video on the timeline then apply transitions or a cross fade I can have events not exactly dovetail to each other on the timeline so the result is I'm introducing blank or empty frames which depending on their length may or may not show up on the timeline unless you zoom in real close. Once you do, you may discover one or more blank (black) frames which you then need to cut out.

This could account for either a fraction of a second flash appearing when you view in the preview window or transations being dark if they hit one or more "blank" frame.

SonyEPM wrote on 8/27/2001, 1:20 PM

It is possible that the MPEG source material lies at the root of the problem. With MPEG2, not all frame data is in every frame(part of the MPEG compression scheme- why draw data twice) but this can lead to trouble if MPEG is used as a source file for editing.

So I would suggest two things if this is a constant problem.

1) Capture something other than MPEG. I know this is a major architectural change for you, but the fact is that MPEG1 or 2 is not in any way an ideal source format. If you can capture as avi (better yet as DV avi) your editing experience and picture quality WILL bet better, even if MPEG is your destination format.

2) convert your source media to avi, (NTSC DV template), do your editing, then render to the final format.

Yes, neither of these is ideal. You might be able to selectively convert problem areas and avoid mass conversion and all the time that involves. With VF2, you can do a "replace" in the media pool once you've converted problem source material and keep all of your current edits.

Also: ALT+arrow left or right key moves the cursor 1 frame, (based on project frame rate).

crombej wrote on 8/27/2001, 9:30 PM

I think you're right about the "sub frame" edits with MPEG being the problem! I tried it in AVI format and couldn't get it to show up.

That brings up a good point. Why don't they switch the left/right key to be frame -/+ and the ALT left/right to be pixel -/+ ??? I mean what good is "pixel" shuttling anyway? Heck it's frames we deal in right? I respectfully submit that key assignment be swapped in the next version! :-) At LEAST make it a preference! :-)

Also, I wish that there was a keystroke combo that always moved forward/back one SECOND at a time. If you zoom in exactly the right amount, then you can do it (because the grid inc is 1 second), but I'd rather just know that ctrl-alt-right will skip forward a frame at a time. I'd just sit on those keys and when I got close, release the ctrl-alt and left/right shuttle to the right spot. I admit that once my hand is on the keyboard I'm loath to go to the mouse unless it's really necessary. Again, maybe a preference?

Regarding the discussion about MPEG not being lossless and AVI being the way to go, it doesn't look like the size will be "too" bad if I go with all DV-AVI format for my edits and then render to MPEG for output to my video card / analog tape.

Maybe one of you format people can answer these two questions?

1a) First of all, I can't capture to AVI, only to DV MPEG (720x480) at a max 10mbps video and 384kbps audio which I think is pretty good? So, when I use VideoFactory2 to do a MakeMovie to convert this to AVI format for editing, I choose AVI and then, what? NTSC DV? Is that a lossless best quality setting? If so, how come I can go to custom settings on that and it has 4 video rendering quality settings (draft/preview/good/best)? Plus all those codecs?

1b) Which format should I choose? I want the equivalent of a compressed LOSSLESS setting, like a run-length-encoding or something don't I? I have 18 different settings in the Video Codec dropbox. Someone said something earlier about YUV9 or something? That's not on there. Is that the best compression while still allowing rapid speed for a decent editing experience? Where do I get that?

1c) After knowing the answer to 1b, should I do another render in say 320x240 size again to give improved editing performance and do a replace when done or isn't this necessary with the recommended AVI format in 1b?

2) (Thought I'd never get to question 2 eh?) What is the effect of repeated MPEG2 DV re-compressions at say highest quality 10mbps video / 384kbps audio? For example, if I took VHS and dubbed it twice, I'd say it's 20% worse, three times maybe 40% to 50% worse (quantitatively speaking of course). What about with this MPEG stuff? Can it survive 3 generations without noticeable deterioration?

Thanks a bunch for all the great replies. I hope others can learn from these answers as much as I am!

John

wvg wrote on 8/27/2001, 10:07 PM

Doing repeated renders of any mpeg will only further detorate the quality on each successive rendering much like doing the same thing with a JPG still image.

Since for now all my projects use mpg as the source file I first render to AVI to "arrest" the quality at the level. I do all editing, applying filters while still as a AVI. I hang on to the the AVI until I'm happy with the result then render one last time to whatever final format I want. It eats up tons of hard drive space. A typical 30 minute vid will hog about 22 GB space.

crombej wrote on 8/28/2001, 1:23 AM

Thanks for the reply.

Again I ask, "When you render to AVI, WHICH CODEC do you choose?" Why?

Which is the best (lossless?) with smallest size and best speed/editing decompression?

I think I asked these questions in my last post among others?

Thanks again!
John

SonyEPM wrote on 8/28/2001, 8:57 AM

The DV format is not lossless, let's be clear about that. The only truly lossless compression that you can use in Vegas/VF is NO compression at all, but the problem is that uncompressed 720x480x29.97 video takes up about 85MB per SECOND and almost no system is going to reliably play that back, render times are huge, storage is a problem etc.

If you work in the DV avi format, video is approximately 3.8MB per second, you get full framerate playback or straight cut DV footage, and you get NO degradation in successive render generations if you use straight cut video- it's purely a data copy to a new file. If you add video fx, tiles etc, of course the video gets recompressed (you are changing the frame data) but the DirectX8 DV compressor is pretty good and is widely used by many people in the industry. Stick with DV as a source format whenever possible- you won't be dissapointed.

Finally, if you are buying new hardware for video editing and wondering what to get, get a generic OHCI compatible 1394 card (such as the Pyro or SIIG), AND the Sony DVMC DA2 media converter. You'll get great results, excellent image quality, you can pipe just about anything in/out, no driver hassles... that's the setup.

wvg wrote on 8/28/2001, 8:58 AM

You asking me John?

If you are, I use the video for Windows NTSC DV template as a first step which seems to be lossless. Since I have several huge drives I don't care about size or speed at this point, I'm trying to tweak out the best quality from mpeg source files. Once I'm happy with the editing, filtering, only then do I render in the final file format which for me usually is either mpeg (so I can burn to CD) or RM at 3MBPS if I'm going to keep for display on my computer.

All the preliminary steps I do only because I'm starting with compartively low quality source files which I assume most aren't burdened with.

As far as the size verses quality issue, they go in opposite directions. If you want max quality, you are going to end up with huge files. If you want smaller files you are going to sacrifice some quality. Its a tradeoff we all have to live with.

As I said in another thread I was very impressed with the quality of Real Media (RM) but only if you ramp it up to 3 MBPS. The rendering time is slow torture taking me over 10 hours to render a 29 minute video. The resulting file was 682MB from a timeline containing a little over 22GB of AVI files.

The RM quality was excellent. The only problem is ASAIK you can't do anything with this file format like have it play off a set-top DVD player which if possible is my goal for most of the videos I'm doing. Still it plays great off a PC.

discdude wrote on 8/28/2001, 10:56 AM

SonicEPM said:

"With MPEG2, not all frame data is in every frame(part of the MPEG compression scheme- why draw data twice) but this can lead to trouble if MPEG is used as a source file for editing."

Would using I-frame only MPEG compression (in essence, MJPEG) solve the problem. I remember in the early days of MPEG-1 video editing, you couldn't edit anything but I-frame only MPEG.

I'm rather surprised that Video Factory doesn't have an option to make I-frame only MPEG.

SonyEPM wrote on 8/28/2001, 11:26 AM

Q: Would using I-frame only MPEG compression (in essence, MJPEG) solve the problem?

A: Yes, it would likely solve the missing frame info problem. Capturing I-frame only MPEG does not mean that MPEG instantly becomes anything more than a barely passable source format.

crombej wrote on 8/30/2001, 12:14 AM

Well, it looks like I'm going to have to render all my video to AVI before doing my final MakeMovie. I guess it won't be too bad - 13gb / hr which is what I figured on.

My question is, do I just choose AVI and then NTSC - DV and that's it?

What about all those options if I choose Advanced Render, then Custom?

I have like 18 codecs available. Is there a lossless run-length-encoded type of choice in there, or is the default NTSC-DV / AVI format lossless?

I know that it really just "copies" the frames unless you're doing a transition in which case it has to re-render. But my question is, since the transitions are a small proportion of the actual video, can I make those lossless?

Which of those zillions of codecs to choose? I tried the LEAD and JPG ones and you have to pay for them. Are they the best?

One final question please? Is there hardware I can buy to make all this rendering run at real-time? Do the Pinnacle 500 or Matrox cards do that so that all editing and rendering is real-time? I'm looking at 9 hours to render 1 hour of video on my P3-800 w/512mb of RAM. Too slow! :-)

Thanks much!

John

SonyEPM wrote on 8/30/2001, 8:42 AM

do I just choose AVI and then NTSC - DV?

yes.

crombej wrote on 8/30/2001, 10:36 AM

Ok, Houston... We have a problem!

I spent 9 hours making the AVI / NTSC - DV movie and it rendered at "half size"!

My original says:
Video 720x480, 32 bit, 29.970fps... MPEG2

I simply dragged it onto the video/audio timeline, did a Make Movie, Chose AVI then the NTSC-DV template and started it up. My new AVI file says:
Video 720x480x24 29.970fps... DV

HOWEVER, when I open it up in media player, it only opens up 320x240 (or so) and if I zoom to 200% it LOOKS pixelated. What did I do wrong? Please help!

Thanks,
John

SonyIMC wrote on 8/30/2001, 11:41 AM

This is a bug with the media player and DV files. Rest assured your file is 720x480,29.97fps even though mediaplayer reports half size.