FCP6 - "Smooth Cam" Processing times?

toddmcm

toddmcm

Carbonite Member
Wow, FCP6 really is a nice app. I keep finding new stuff that is just awesome.

Smooth Cam is probably my favorite new feature.

I'm wondering who's using what kind of system and the "analysis" times you are getting.

Takes my dual 2.5 G5 about an hour to process 10 seconds. I threw a 48 second clip in and yup, took 5 hours. Then of course you have to render the clip.

So it's basically 10sec = 1hr on my machine.

It's cool that it processes in the background... but I can't help but wonder if it could analyze much faster if it wasn't just a background process. Anyhow, what kind of times are you guys seeing?
 
5 hours to process a 48 second clip!?!?!?!? Someone please tell me their getting better times than that!

Toddmcm, what kind of footage was it?
 
Ymmv- keep in mind that Smoothcam will analyze the entire piece of media. So for example if you captured a 10 minute clip and you drop a 5 second piece of it on the timeline, it will analyze all 10 minutes. So if your work consists of long captures, do an export of the section you need *before* you start smoothcam on it.

-Noah
 
From what I saw at the demostration, the SmoothCam looked straight out of Shake. I love using it in Shake and look forward to seeing how well they adapted it to FCP.

I agree with Noah. If you're looking to just smooth out a small portion of the clip, you might want to cut it up. I remember doing a long clip in Shake and it didn't take anywhere near as long as it seems to take you guys in FCP.
 
Noah's got the answer... odd that it does it this way -- but also good.

One draw back from the shake method is that it gives you back STRAIGHT clips... not transitions.

This way - even if you transition into and out of the clip you'll still have the smoothcam applied.

Yes - it takes a bit of time. But who captures full tapes anyway??????? (hee)

CaptM
 
No way SmoothCam comes straight outta shake. At least i dont think so. If you detach the filters tab, youll see that smoothcam is an fxplug, developed by a third party...

And the times of shake's smooth cam are way better than that. I mean waaaaay better.
 
I was playing with it today-- I had a 1 min 20 sec clip and it took 12 minutes.

I have a dual 2.6 mac pro.
 
I was messing around with SmoothCam with some HDV footage (as I didn't have any of my DVCProHD material with me) and discovered some really unusual things.

I took a 10 second clip, analyzed it, then rendered it. When the render was finished, the SmoothCam plug-in PUSHED IN on the footage. It thought the not-so-tripod-like camera movements were too much and just zoomed in past them, I guess. I don't know.

I'll keep playing with it, but it just seems weird.


EDIT: I found out the culprit was the Auto Scaler. Well, that and the ridiculous camera moves I thought this plug-in could overcome. Silly me.
 
Last edited:
Interesting indeed. 1 minute 20 second long clip and it analyzed in 12 minutes????? Something is wrong, obviously. Or is it just the G5 vs MacPro???? Damn, if there is that much of a difference I'm buying a new machine.

The footage I was "smoothing" was surfing footage that was hand held. It works wonders on the footage but takes forever. The clips I was testing on were full clips, the one was ten seconds, the other was 48 seconds like I said.

1 hour per 10 seconds regardless of footage on a dual 2.5 G5

I plan on testing in shake as well. We use shake at the studio, but I've never used it for anything other than compositing our 3D shots together.

Now I'm VERY curious what my times will be in Shake.

I'll report back over the weekend. I don't even think I have Shake installed at home but I do plan on testing.

Thanks for the feedback.
 
The way smoothcam works is to analyze x-y movement and move the frame to counter it. In order to do this without revealing the edge of the frame on a given shot, it has to be scaled up a certain amount. It tries to do this with the least amount of scaling possible to avoid image degradation. So the higher resolution the footage the better the results will be.

Noah
 
Lets be honest, this is not an ideal implementation. It seems this was 'tagged-on' for the release. Its a great tool, but ultimately it needs to just work on the edit length.

Exporting, re-importing and cutting over a shot is a solution, but its not progress in terms of workflow. Clearly they need to integrate it with the FCP engine so it can pick up only the length of a cut.

Its true that clips will often be captured as short shots, certainly more so in slated situations, but it seems to me this type of tool would be most used in docu type work where there is more chance of camera shake issues and they are not known for their short shots :-)

Just my 2 cents.
 
Agreed- it likely comes from the Shake mentality of not being applied until clips are fully edited. I'd recommend using the Provide Feedback option in the Final Cut Pro menu. Be short and sweet- it does get read by an actual human being. :)

Noah
 
10 second piece took two hours. DVCPROHD at 1080p/25. Masterclip's duration was 4 min, 10 sec. Looks like SmoothCam really "analyzes" all whole source clips…
BTW: iMac CD 2.0 GHz/2 GB RAM, fast drive
 
Seriously, this is making me crazy. Why on earth is it taking so long?

No matter what I try, it's 10sec=1hr.

In shake, it's really NOT much faster to get equal results. Sure, you can dummy down the shake node and get it to process faster but the results are nowhere near the same. Once you turn the settings up in Shake to get great results the time is actually longer.

Seriously considering buying a new machine. I want to sit down and test on one before I buy. I wonder if Apple store will let you bring in a CD with a clip to test.

I'm having a really hard time believing the 1minute 20second clip analyzed in 12minutes. Was that with HDV footage or DV. I tested with DV and it's light years faster.

Anyhow... I'm trying to get a more definitive test going here. I'll report back.
 
toddmcm said:
Seriously considering buying a new machine. I want to sit down and test on one before I buy. I wonder if Apple store will let you bring in a CD with a clip to test.

I'm having a really hard time believing the 1minute 20second clip analyzed in 12minutes. Was that with HDV footage or DV. I tested with DV and it's light years faster.

Anyhow... I'm trying to get a more definitive test going here. I'll report back.
Click to expand...

Honestly, I took a small clip and began to analyze it in FCP 6 on my dual 2.3ghz G5. At about a third of the way through it's process, I thought to try it on my 2.0ghz MacBook Pro. Because I exported the clip I was testing with the G5 (to avoid analyzing the entire 10 minute clip), I used the same clip on the MBP...

I moved the clip to my MBP via wireless, imported the clip, and began to analyze it. Once it finished, I rendered it.

It finished 10 minutes before the G5 finished and I had started the MBP when the G5 was 1/3 of the way through, AND, the MBP clip was rendered and ready to play.


Needless to say, once the quad procs come down a bit, I will definitely be upgrading my G5 to a MacPro.
 
toddmcm said:
No matter what I try, it's 10sec=1hr.

In shake, it's really NOT much faster to get equal results. Sure, you can dummy down the shake node and get it to process faster but the results are nowhere near the same. Once you turn the settings up in Shake to get great results the time is actually longer.

Seriously considering buying a new machine. I want to sit down and test on one before I buy. I wonder if Apple store will let you bring in a CD with a clip to test.
Click to expand...
I just test a couple of clips with a MacPro w/4mg RAM. I used 20 sec clips and they varied a little from 10 to 12 mins each. it seemed to be related to the amount of camera shake. Both were pretty shaky.
I can't imagine 10 sec clip taking a an hour.
I'll test these same clips on a Mac Book Pro later today.


David
 
I've now gone as far to shoot the most BASIC footage possible. One flat colored object against a plain white wall in my house. It still is about 10sec=1hour.

Dual 2.5 G5 / 3GB RAM / 1TB eSata RAID

NOW, the interesting point that Nik brought up. Is it using the GPU as well? Or entirely? That would better explain this as there isn't THAT much a difference in full blown CPU calculations between a dual 2.5GB and a CD2 MacBookPro.

It would make sense that the measly 9600XT in my machine is the problem, or that combined with the older processors.

One thing is for sure, it is NOT the drives. I still have not had a chance to test one of my clips on a MacPro.
 
Sooo Uhhh, put my Dual 2.5 up for sale on ebay, went to breakfast, returned home and it had sold with buy it now.

Shiny new 8 core on it's way. I figured that should help, and I can use the 8 cores on the windows side with 3DSMax. STOKED!!!!!!
 
You must log in or register to reply here.
Back
Top