excessive encoding time?
 

I'm editing 1080i60 HDV footage in FCP 5.1.4 on a Macbook Pro 17" 2.16GHz. I've got a 4-minute sequence that I'd like to encode to a 40MB or so size file for the web.

I tried exporting to quicktime conversion, using H.264 and a frame size of 600x338 ... and it takes close to two hours to encode.

I also tried exporting to compressor, using one of the H.264 presets, and that also takes about two hours.

Two hours to encode 4 minutes -- is this normal, or excessive encoding time?

Should I be using a different method to create a clip for the web?

Thanks.

Dave,

Unfortunately you're experiencing the typical HDV workflow issue, which is that because HDV is a Long-GOP codec it requires a great deal more computing power to work with than an "i"-frame type codec.

At issue is that HDV/long-GOP is a "group of pictures" similar to the GOP structure in MPEG-2 encoding for DVD. Simply put, it means that one group which in HDV means anywhere between 12-15 frames are actually being handled as one single packet whereas HDCAM, DVCPRO and others are frame-independent - each frame stands on it's own and is not tied to adjacent frames to be considered complete.

Think of it in physical form - and you'll get the idea of why your system is having such long render/compression times: Let's represent each single frame of video as a 10-pound iron weight. Moving that single frame - or weight - would not require a lot of effort on your part, you could easily pick it up and carry it say over to a desk near you or across the room. However, if you had to move 15 of those "frames" together in one move as a group now you're moving 150 pounds all at once! That would put a serious strain on just about anyone - unless they were either a bodybuilder or worked out with heavy weights routinely.

This direct analogy is the same for your computer; HDV requires such intensive work from your system - especially a laptop which has limited CPU, HDD and RAM power - that it is quite literally being strained to the limit when being asked to render or compress HDV clips - or any other Long-GOP structure format.

Unfortunately there is no cheap or easy way around this issue; there is a device from Convergent Design which converts Long-GOP into any "i"-frame codec but it's not cheap and requires using either a KONA or BM card.

Thanks, I was afraid that was the case. I suppose it is time to look at one of those quad core 3 GHz Mac Pros, eh?

H.264 is also a very time consuming codec to encode so that's another facet of your problem. You might try using a different codec if you are going to be putting things up on the web often. I've found that the Sorenson 3 to be pretty flexible and it is compatible on a wider variety of machines than h.264.

If you want to try it out the following settings would be a good place to start:
Sorenson 3
Keyframe - Automatic
Data rate - Automatic
Quality - Med
Applelosses codec for Audio

Then just tweek as needed.

-A

Thanks Andrew, I will try that ...

If you don't mind the limited options there is this little USB H.264 accelerator from Elgato.

http://www.elgato.com/index.php?file...ts_eyetvturbo#


-A

I am afraid the result after the investment would be still disappointing. I doubt the render speed will go beyond double, means your 4 minute clip will render in 1 hour instead of 2 - after spending the $$$. Now think of a 40 minute clip instead of 4, still 10 hours (well, not almost a day perhaps).
The big issue is that you see your final result after a very long time (you cannnot do other stuff in the meantime, otherwise your rendering is even longer, etc)

One may think thats fine, edit during the day, render at night. Maybe, but....

I had my HDV project on for 3 consecutive days to render, just to realize the settings I used produced a soft image... So I had to restart the entire project render again and again, while became sick of HDV editing.

I read forums instead of working, but no one seems to give a straightforward abswer about a workflow that:

- provides with no (or minimal) loss in final output (at least vs the original HDV quality)
- works with acceptable render times
- and list of hardware/software that is necessary to ingest the footage and/or do the output

Anyone here with a comprehensive HDV workflow for Final Cut setup?
Capture cards, ingest in which format, edit in what timeline setting, using Compressor or what else for output for high def final materials (e.g. blu ray)

Thanks a lot.

There are lots of options to enable users to speed up there workflow but most of these will cost you. We are now working on files with much more to information to calculate and the basic hardware of yesterdays Mac/PC will struggle.

I will be upgrading my setup soon, as I have the first generation Dual 2.0 G5 Mac. I cannot work with long HDV sequences as it is too slow for me.

I capture as .m2t convert to 720 25p AIC, whilst it's not as quick as DV it allows me to try things out that I would not do with HDV due to render times within FCP and outside in Compressor.

Capturing .m2t and converting to 720 may seem like defeating the object of a fast workflow, but these tasks are preformed in the background whilst I work. And pay for themselves later on.

720 AIC goes to H.264 or FLV's allot faster than HDV.

I will always try 10-30 seconds of my clip into the final webcodec of choice just to make sure. It's very easy with compressor to select the in and out point for this test.

I recently upgraded to an 8 core with 5gb ram. My rough unscientific comparison with my Dual 2 Ghz G5 on a 1 hour 20 minutes HDV sequence...

Render time
Dual 2Ghz = 22 hours
8 core = 5 hours

Compressor test
DVD best quality in compress for previous sequnce after export...
Dual 2Ghz = over 3 hours
8 Core = less than 1 hour

Needless to say, working native in HDV just requires faster machines...

EDIT: These tests done with FCS 1. I have read elsewhere that Final Cut rendering as well as compressor 3 can be much faster due to taking advantage of all 8 cores.

Cheers,
Scott

Thanks Scott, what you are saying is pretty convincing. I am hesitating over an 8 core purchase, not solely due to the price of the macpro itself but the additional costs, too.
I have an e-SATA raid PCI-X card, a TC-electronic effect card and a MOTU sound card - all in PCI-X flavor.
Switching to a macpro therefore would at least cost a PCI-X to PCI-e box, thats about $1500 and a video card perhaps ($250 for Intensity if not something else).
To fully utilize ProRes (as in capturing in ProRes), I may need to go for a HD-SDI card instead.

Now here is the dilemma:

what if I just keep the dual 2Ghz G5, get an AJA IoHD (then I would need just a Firewire connector) and ingest/capture everything in ProRes, edit in that?

How does ProRes work in terms of rendering times?? What I have seen was that rendering times are not really good vs HDV. (please correct if I misunderstodd something)

Does it make sense to capture HDV footage into ProRes, working all the way in ProRes 4:2:2 (color correction, etc)? Apparently the quality wont be better than the original HDV - but in the same time it wont be worse, plus I will have more room to play with color, effects, etc.

And...maybe I could still keep the G5 for the time being, with the need for firewire only in AJA IoHD. (My Sony V1E has HDMI to connect it to the Io)

Having said that, it is clear that in the HD business I cannot avoid investing in a faster computer - I just expect some better Macs coming out very soon (with the new OSX perhaps), so I just try to manage things until it happens.

Also weird that blu ray is not supported by FCS2...it makes me think some major upgrade will be coming.

Thanks for any advice. If that helps, I plan to switch to XDCAM EX once it will be out...

I will never be convinced that converting HDV to a codec (prores) that takes up 5 times more space and doesn't render faster is beneficial. Especially if the delivery medium is still eventually SD DVD.

I debated faster computers coming out too...truth is I don't believe there will be any major improvemtn over the 8 core this year. You may see small changes, maybe even a new MacPro enclosure....but I doubt it will contain signifigantly more horsepower.

Cheers,
Scott

Scott,

in your example project, have you used color correction, filters and transitions or just plain cuts?
If yes in what extent?
I tend to use quite some color correction and some dissolves. It slows down HDV processing a lot. I wonder how this affects the 8core performance.

Thanks a bunch.

Tons of CC, effects as well as layering (using composite modes)...Plain cuts have much more reasonable render times on my G5. With no CC, effects or layering (compositing?) I'm gessing the 8 core chews thru it in even less time...

Hey, ya wanna talk render time...I just read that it takes 38 hours to render a single frame of the new Transformers movie...What a bunch of whiners we are.. ;)

Cheers,
Scott

I also use this method, as I found HDV to slow expecially for long clips. I also found that I had a greatly reduced number of realtime effects when using HDV on the timeline. I have a G5 dual 2.0 with 4gb ram. AIC is just faster.

Drew

Encoder solutions
 

There are some encoding cards in the pipeline that will offer a significant performance boost. Among them there is a product from Aspex Semiconductor called the Accelera 2500/3500 which will speed AVC h.264, VC1 and Mpeg2. The 3500 ($9000) will encode SD h.264 5x real time speed and HD in .5x speed. The 2500 can only work with one stream at a time with about 50% of the horsepower of the 3500 for $4995 list price.

While this is a great deal of money, it is a bargain compared to what you would spend on a cluster of servers to match the performance; not to mention the simplified workflow.

They have officially announced support for Apple OS X and will be shipping soon.