Is 869MB about right for 12 minutes?
 

I've been successfully editing 5D mk II footage using Cineform NeoScene transcoded files in a 1920 x 1080, 30FPS, Square Pixels Premier Pro CS3 project. I've output the timeline via DebugMode FrameServer at 1920 x 1080 pixels, 1:1 PAR, to TMPGEncXPress to a standard DVD MPEG file, using the provided DVD MPEG template (720 x 480 pixels, widescreen).

The resulting file is 869 MB. This is for a 12 minute clip. I've output significantly longer video clips from my Canon XH-A1 using the same method (but only 1440 x 1080 pixels), to the same DVD MPEG template in TMPGEncXPress and the file sizes of the output files are a lot smaller (less than about half the size).

The following attachments show my settings in Premier Pro CS3 and TMPGEncXPress.

Do you think my settings optimal for DVD output? Any suggested changes?

Julian

On the top left setup page.... you might want to try and change the timebase to 30 instead of 29.97..

Then on output use the AME... the conversion will put the timebase at 23.97 on the output.

This way, the Native 5D footage of 30p is ingested at 30P but converted upon output to DVD...

Oops! Yeah, missed that one! I have 2 installations... it's correct on one, but obviously incorrect on the other! Thanks for catching that.

I've used AME to create the DVD files before, and found the quality is less than stellar. That's why I went with TMPGEncXPress. A 600 MB DVD file created through TMPGEncXPress is frequently only about 300 MB through AME. The quality (judged just by looking at it) isn't half as good, but it's obviously not as good.

I usually output the MPEG file through Encore, which does the conversion for me and produces pretty good results.

I'm just not used to seeing such large file sizes for DVD MPEG files which are only 12 minutes in duration. I know there are more pixels (1920x1080 instead of 1440x1080) and more frames per second, but wasn't expecting such a difference.

Julian

Julian, your source parameters (1440 non-square vs. 1920 square) shouldn't have any impact on final file sizes. In fact, even an SD source should yield the same size output, assuming the runtime is the same... since you're using CBR. You might want to double-check those figures. CBR should be CBR... otherwise, something's up with the encoder.

Also, you're really pushing the theoretical maximum of DVD-spec. on video bitrate. Yes, 9.8mbps is the published max., but many players will choke on output long before that max. You might outta go with VBR, 2-pass and really let the encoder push your quality up in scene areas where you really need the bandwidth.

And I know we've bemoaned the bad AME downconverts in the past with CS2 and CS3. But I've had some pretty amazing ones now that I've upgraded to CS4. In fact, for most of my projects lately I've been able to downconvert 720p Windows Media files to SD DVDs without even re-rendering from the original PPro sources. I have moved to 24p projects (and not 30p), which may explain some of it (fewer frames per sec. to encode)... but CS4 might have resolved it. Looks like I'll have to do some more testing...

Oh... and I've always heard that it's wise to uncheck "optimize stills" in PPro rendering. It's said (since version 1.0) to slow things down.

HTH,
Brian

Also of note... I have heard its best to ingest your HD footage into PPRO, then output the footage to SD without any edits to the footage... then ingest the SD footage and work the edits on the SD footage... this would be AVI outs and ins...

It's a minor point, but according to Canon Support, the 5D2 video is 29.97 fps. The 30 fps is an error in MOV file metadata. If you can't trust Canon ... 8-O

If the Metadata flags PPRO with 30P but its acually 29.97P then PPRO will force a render
if you use 29.97P...

this could be a pretty big issue....

The 5D2 has much, much more aliasing artifacts than the XH A1. To the computer, aliases are indistinguishable from real detail. Aliases are preserved (or worsened) when downsampled to lower resolution (DVD). Since the Cineform compression automatically varies with the amount of detail, it's no surprise that 5D2-originated footage requires a much higher bitrate.

Hope that helps.

It could be a serious problem if you are mixing the 5D2 footage with non-5D2 29.97 A/V tracks in the same project. Else, it ends up being a minor change in speed, which is not noticeable. The issue really hinges on whether various NLES use the frame rate metadata or not. I have done this experiment and posted it elsewhere. I changed the frame rate entry in the MOV file with a metadata editor, from 30.00 to 29.97 and it didn't seem to make any difference in either Vegas Pro 8.0, Premiere CS3 or CS4, as long as I used a 29.97 timeline. I won't say it was an exhaustive effort and I wouldn't extrapolate the result to all NLEs. I could be wrong on this, but then again, Canon Support could also be wrong. As to the latter, I tend to doubt it, since I have read a number of other message boards and forums where people are claiming that the 5D2 shoots video at 29.97 fps.

It would make a difference if you are making some fairly large edits... because the NLE
is going to force render the time base and attempt to get the large edits in at the same time....

If the NLE doesn't have to work on the time base conversion then the NLE can handle the
edits in a much more controlled way...

This is another reason to ingest the raw footage, then export the footage and then engest
the footage again prior to performing the edits....

pretty much another benefit of using Cineform for the re-wrap