is Neo Scene even needed in Premiere Pro CS5?

[Ed Kishel]

Was about to upgrade my neoscene for my new CS5 machine- when I ran accross a video on Adobes site that states transcoding in not needed:

Premiere Pro CS5 Feature Tour - Staying Native or Going Intermediate? Transcoding and Premiere Pro CS5 | Adobe TV

I knew that if your machine can handle native formats, that it wasn't neccessary to convert- but I did anyway to get the 4.2.2 colorspace "upgrade" for post work. But Adobe goes on to say that PPro does this automatically to native formats.

So if mercury handles native formats and does the 4.2.2 on its own- why do I even need Neoscene anymore?

[Stephen Armour]

Ed, there are a whole bunch of "IFs" in his presentation:

IF you have a cutting edge machine
IF you only stay within Adobe CS5
IF you have an accelerated GPU
IF you are ONLY a standalone user, not in a workgroup workflow
IF you mostly do basic editing and not multi-generational passes
IF you don't do much chroma work
IF you really don't need output master's for multi-target type output,

THEN...what he says is true.

Remember, he's an ADOBE CS5 EVANGELIST, so his story is carefully crafted to tell things that are true...to a certain extent. It's what they DON'T tell you that is important for you to know before trying to go their route!

There are many reasons why CF is still more than valid for today, and many of those reasons are what they are NOT telling you.

[Nathanael Iversen]

Because the footage looks better, takes more tweaking before showing artifacts in my experience. Very nice products. If you are just butting footage from one camera together, maybe not so needed. I have multiple cameras, (including different brands) to integrate.

First Light is very nice - it is like having Lightroom for still picures. I find it very easy to use.

[Robert Young]

Without doubt, on my CS5 system I can do a simple cut & stitch edit in native AVCHD and burn it out to a pretty nice looking Blu Ray.
However, for more extensive editing involving color correction, effects, graphics, and etc., transcoding to a variety of delivery formats, the Cineform codec is far more durable, and will reliably produce a finished delivery product that maintains the image quality of the original footage.
The underlying problem really has nothing to do with Premiere Pro, but rather the fact that AVCHD is a highly compressed, "lossey" acquisition codec that does not tolerate extensive editing very well.
In comparison, the "lossless" Cineform codec is designed specifically as an editing format that nicely survives the editing process.

[Stephen Armour]

How about with your EX1 Bob?

[Bill Bryant]

I started out editing natively in CS5 but I like the Cineform experience much better. I hate all the hard drive space I am now using, but navigating the timeline is silky smooth compared to my native files. And First Light is just wonderful, especially on long form multi-camera edits.

[Nathanael Iversen]

Originally Posted by Bill Bryant

And First Light is just wonderful, especially on long form multi-camera edits.

Very true! It was very easy to match up very different takes on the same event.

[Ed Kishel]

Originally Posted by Nathanael Iversen

Very true! It was very easy to match up very different takes on the same event.

I'm talking about NeoScene- this does not include first light so that isn't a factor for me. I'll put my old version of Scene (last build before v5) on the new CS5 machine and I will see who slows down first after adding some effects. My guess will be AVCHD- but if it isn't a night and day difference, I doubt I will upgrade. All video related activities stay within CS5 so I wont need to render until final output- so I won't need multigenerational passes for different applications. I would use bridge to send the same timeline between programs (PPro to AfterEffects back to PPro to Encore as many times as I want without rendering).

Hey CF, how about a First Light SE ( a lite version of First Light ) that comes with NeoScene. Charge $150 and you got a winner :)

[Robert Young]

Originally Posted by Stephen Armour

How about with your EX1 Bob?

Maybe the EX/XDCam codec is a little more robust than AVCHD, but honestly, I never tried any real editing in the native format. I always use CF.
The systems are so fast now, I never really think about the conversion to CF as an issue. I dump the original footage on an external drive, fire up HDLink to output CFHD to the RAID and go get a cup of coffee.

[Robert Young]

Originally Posted by Ed Kishel

I'll put my old version of Scene (last build before v5) on the new CS5 machine and I will see who slows down first after adding some effects. My guess will be AVCHD- but if it isn't a night and day difference, I doubt I will upgrade.
All video related activities stay within CS5 so I wont need to render until final output- so I won't need multigenerational passes for different applications. I would use bridge to send the same timeline between programs (PPro to AfterEffects back to PPro to Encore as many times as I want without rendering).

Ed
Your logic certainly sounds reasonable, and, since you already have CF, you are in a position to put it to the test in the two arenas you are looking at: the editing experience, and the final delivery format image quality.
You could put together 2 identical short projects, one as native AVCHD, the other as CF- using the same clips, maybe just 3-4 minutes long. I would trim the clips to about 10 sec duration- long enough to absorb image detail when you do the final comparison for image quality. Add identical titling & simple animated graphics (a quick round trip to AE using dynamic link), some animated transitions, treat a few clips with effects of one sort or another, do a small global color correction to each entire timeline (maybe just boost the color sat a twitch), do a PIP or two, etc.
Then observe:
1) the editing experience- compare the preview playback quality, scrubbing, etc. Compare the speed related to the editing process. And so forth.
2) Final image quality in various delivery formats- Follow your normal workflow and output both projects maybe to BR disk, DVD, and Flash for web. Play the disks on HDTV, Flash on Adobe Media Player and look closely. If you can't see any difference, then it doesn't make any difference :)

[Ian Lewis]

For what it's worth... I've been doing a lot of XDCAM recently, too. I always output to a final Cineform master file from which to make copies, DVDs, downres etc, but do sometimes work with the native files copied straight from the camera cards.

Honestly in basic editing I don't notice any difference, but i still wouldn't like to do too much effects work on the mp4 files. In terms of timeline performance, I also find no difference, working with Desktop playback in both (so far I can't get CF playback to switch properly between source and programme monitors to an external screen).

Of course, encoding the rushes to Cineform also gives you a backup - especially if you do it to a different disk from the XDCAM originals - and, as we know, data doesn't exist unless it's on two separate hard drives.

Also, as it happens, I use CatDV for media managing a growing archive of camera material, and CatDV is happy with CF files, but doesn't like the XDCAM mp4 or MXF files.

So I can understand why you might not want to buy CF if you don't have it, but I still think it's worth it if you do - and certainly I think it's important for the best quality and most robust final output. One of the reasons that stations like the BBC and Discovery are so picky about what they'll accept in tapeless production, isnt just the initial straight-from-the-camera picture quality, but the robustness of a picture that's going to be uncompressed and recompressed repeatedly, not just during post, but on its journey from playout on transmission to your telly at home. And that's where Cineform starts being worth it.

Ian

[Adam Gold]

Originally Posted by Robert Young

You could put together 2 identical short projects, one as native AVCHD, the other as CF...

I'm actually doing just that even as we speak; I have timelines rendering now and will report back soon.

Okay, so I have some preliminary results, and to cut to the chase, in some situations not only is Cineform unnecessary, but it can even slow you down.

Now all the caveats while everyone howls:

I do in fact meet nearly all the criteria Stephen posts above, although I wouldn't exactly call my machine cutting edge and I do output to a variety of formats. Cineform is the first to admit that they have moved from a simple accelerator to a much more sophisticated suite of tools. But I don't need those tools, and maybe not everyone does. I was, and am, only looking for the smoothest and fastest editing/playback experience. And for that the native CS5 projects outperform Neo4K, which is what I used.

Here's what I did and the results:

I took a roughly one-hour AVCHD clip shot on my XR520v and imported it into a standard AVCHD project in Premiere, then transcoded the clip using HDLink at High quality and imported it into a Cineform project. I applied two effects to the entire timeline, one GPU-Accelerated (Timecode) and one not (Auto Color). I then rendered the timeline for preview using both GPU acceleration and software only in both projects. I did not do any transcoding or exporting.

Findings:
The standard clip took about five minutes to import using Sony's PMB. File size was about 5.52 GB.
Conversion time to CFHD-AVI took 36 minutes, and resulted in a new AVI file that was 73.7 GB
In the Cineform project, Time to render timeline
With MPE GPU Accel 2:30:00 (approx)
Software Accel Only 2:30:00 (approx)
In Native Project
With MPE GPU Accel 59:30 (approx)
Software Accel Only 59:30 (approx)

Obviously, having the non-GPU effect wiped out any difference between using the GPU accel and the software only mode.

So I removed the Auto Color (Non-GPU) effect and re-ran the tests. Now the results were even more interesting:
In the Cineform project, Time to render timeline
With MPE GPU Accel 1:46:00 (approx)
Software Accel Only 1:07:00 (approx)
In Native Project
With MPE GPU Accel No rendering necessary
Software Accel Only 40:00 (approx)

Why using the GPU made the Cineform render slower than software only has me baffled. But I double checked and ran the test multiple times.

In both cases, full screen, full res playback was smooth as butter. There was a difference in scrubbing the unrendered timeline with effects applied; the Cineform scrubbing was smoother. But once rendered there was no difference.

Rendering the Cineform timeline used about 25% CPU and about 4 Gigs of RAM (out of 12 on the machine), and never used more than five cores on the i7 chip. Rendering in Premiere always used all eight cores at about 90% CPU usage and about 9 gigs of RAM.

Obviously this is not a real-world test, but it is apples to apples. Not everyone will use the effects I chose, and everyone's results will likely vary. So all the usual caveats apply.

I love the Cineform guys and their service and participation on this forum is exemplary. I'm still a huge fan. But for me, Neo4K does not provide a benefit that I can see. But then I'm a simple man with simple needs.

[Robert Young]

Originally Posted by Adam Gold

Rendering the Cineform timeline used about 25% CPU and about 4 Gigs of RAM (out of 12 on the machine), and never used more than five cores on the i7 chip. Rendering in Premiere always used all eight cores at about 90% CPU usage and about 9 gigs of RAM..

That's an interesting observation.
When I render out a CF timeline (Neo4K v 5), my Win7 64/ i7 system runs around 6-7GB RAM usage and drives all cores to pretty much the max.
Also, I don't have any exact numbers to offer, but my rendering time seems a lot faster than you report.

[Adam Gold]

I'm sure there are ways to tweak both the HW and the SW to encourage better performance, but I just accepted the default settings for the appropriate project preset for each type of video to keep it apples to apples. And I deliberately tried to stress the CPU and card by applying effects that forced re-rendering of every single frame of a one-hour timeline.

Perhaps someday there will be some documentation from Cineform to help with this, but the Documentation PDF package that comes with the latest Neo4K download is for Neo 2k, from 2007.

[Stephen Armour]

Thanks Adam, for the interesting comparison. I guess that just confirms what we've seen and experienced and what a careful in-between-the-lines reading of Adobe's come-on tells us.

If we were in your shoes, we'd probably not be looking much at CF'd stuff any more either! But as we are in the opposite camp from you (like with others here), there is still great value and long term benefit for us with the CF workflow.

The multiplatform/workgroup/chroma/multioutput/open-ended res aspects of it's intermediate codec workflow, make it more than just "good" for us. They move it into the "necessary" and "quite desireable" camp.

Different strokes for different folks, and it pays to research well these days. CF did you good for a long time, but nothings stays the same in our fast-moving world, that's for sure.

What I see coming though, is a need to stay riding CF's wave for the near-to-medium future, especially as they keep the 3D and extreme-high-res forefronts covered. No matter what we think of the merits of those aspects, they are coming very quickly down the pipeline. CF has shown they can respond just as quickly to these cutting edge tendencies, so certainly warrant a continuing connection for those of us with existing CF software investments. Just the fact that CF's chief engineer's working with MS on Sliverlight tells you to keep tapped in. That kind of connection gives them a considerable view of things down the road.