Personal View site logo
GH2 Cineform Question
  • 76 Replies sorted by
  • @notrons I have read and reread what you wrote, and it seems like what you are saying the post prep utility does makes a lot of sense as far as work flow.

    Check me if I am right here. To put it simply you will take your original footage and convert it into two types of footage with the exact same name and file type.
    1. will be a medium quality footage with 422 color space so that your color corrections look accurate. 2. will be a uncompressed (maximum quality) up sampled 422 color space footage with the same exact name as the footage from number one.

    When you get done editing your project, you will simply overwrite the lower quality footage from number one with the very high quality footage from number two. At that point, your editor will simply apply the same edits to the super high quality footage which will in turn slow your editor down. But this is unimportant because the only time you use this very high quality footage is when you are exporting?

    The next question would be ... is the post prep utility able to process the files as well or better than neoscene or the other cineform versions? Because if I am correct, you could accomplish a very similar thing to what you are doing with those programs. You would just have to process the files twice instead of automatically having the program do both files for you.

    On a side note. I have expertise regarding RAID. In both of your posts you mention RAID, but you mention two different array types. Personally for critical data storage, I prefer RAID 6 because it allows any two hard drives to go down at the same time. This actually occurs more frequently than you would think because hard drives that come from the same batches tend to fail at the same time. RAID 6 also integrates striping but with parity. THis gives you good speed with great backup capabilities. Of course it does require a 5 or 6 or more disc array, and many RAID controllers don't support it. But it is awesome.

  • @notrons

    When you say that the super-bit rate GH2 footage is like molasses on the time-line, are you talking about intra-frame footage? Or long GOP stuff?

  • GH2+cuda=best quality. Plus real time playback of multiple effects for CC. No need for Cineform.

  • @DrDave If you are using Vegas Pro 11 as opposed to current Premiere Pro, it uses OpenCL so it doesn't matter if you are using ATI or NVidia (and your card doesn't have to be on the approved list).

    If you are using Premiere Pro, then an approved CUDA card is required to use the Mercury Engine.

    If you are using Magic Bullet Looks 2.0, then the color correction is OpenCL and benefits from any sufficiently fast GPU, just like Vegas Pro 11.

    If you don't have any GPU, Cineform CC outperforms both of them, since the color correction typically runs in real-time.

    Note that Vegas Pro 11 "seems" to outperform current Premiere Pro in certain benchmarks because of the GPU being used to accelerate more areas (like the "pan and crop"/"Ken Burns effect" processing). I can't say properly myself because the CUDA card I used to try and compare the two didn't work properly with the newest Premiere Pro at that time (even after I found and applied the most popular patch for that purpose).

    @tderuiter Anyway, if you don't have a good GPU, using Cineform is one of the only viable options for real-time color correction.

    If you have a good GPU, Magic Bullet looks can usually take advantage of it in most NLEs it runs on and will perform the same whether you are in a lower priced NLE (like Vegas Platinum 10) or a more expensive one (like Vegas Pro or Premiere Pro).

    But that's just in regards to the color correction itself. As regards .MTS playback, I wouldn't want to start debating GPU accelerated Vegas Pro vs Premiere Pro because I couldn't run the benchmarks I wanted to.

  • @tderuiter Also, the "FirstLight" color correction in Cineform is very cool, but Magic Bullet Looks (and other specialized color correction tools) offer features that it does not. For instance Looks offers additional modules (diffusion filter, two kinds of bleach bypass, telecine net, etc.) that may make it useful even if you use Cineform for FirstLight.

    Also note that even with acceleration, Magic Bullet is typically slower than Cineform. But the GPU accelerated native CC in the newest Vegas Pro (and I've heard in Premiere Pro) are much faster than Magic Bullet Looks.

  • @thepalalias You don't need an approved card. Even with a $35 card --GT240 with 1gb DDR3-- you can layer and layer on the effects and it will play in real time and render like lightening. I have never seen anything so fast. The hidden feature, however, is the superscalar resizing--a must for Ken Burns type stuff or pan & scan with video. It essentially gives you access to the proprietary resizing algorithms which are amazing--also in real time. Compare the quality, you will see.

  • @DrDave No disrespect intended, but I bought a $90 card and returned it when it didnt work, despite being an NVidia CUDA card. It did however do just fine in every other application I threw at it, from 3D animation to Magic Bullet Looks. Im sure it works on a very wide range of cards, but its not universal and it pays to check the specific card first. Thats all I was saying. Also, Adobe did (at least last year) maintain a list of cards that work and there was a hack to get around it. That may have changed more recently.

    Either way, I didn`t get to test that aspect yet, but look forward to it in the future.

    As to the rescaling, I look forward to it. It`s utterly ridiculous the extent to which NLEs have continued to rely on Bicubic rescaling when something as simple as Lanczos3 is far superior (let alone the proprietary algorithims used in some of the other still image programs like Blow-Up, Perfect Reize, PhotoZoomBrowser, etc.)

    Do you know which algorithim it uses? I would be curious to see some comparisons since its an area Im always looking to improve.

  • All of the card issues were figured out ages ago. All you need to do is enable Cuda in the Nvidia panel, and then carefully edit the text file that has the list of approved cards in the Adobe folder. Simply type in the name of your card. If you do not type in the name of the card, it won't recognize it, but if you do, it will. Even works on laptops. All you need is a GT 240 with 1gb ram. You can buy one on eBay for $30, or you can spend a bit more and get DDR5. In my testing, you will only need the DDR5 if you are using loads of effects on the same clip. The scaling, I believe Lanczos2, but it is tweaked somehow.

  • @DrDave I went to read up on it. Adobe info via Todd Kopriva seemed to indicate I would need a GPU with at least 1.5GB RAM to support the frame sizes of the stills I use but that the algo. was Lanczos 2 with lowpass-filtered bicubic. Intriguing. For static rescaling I typically used VirtualDub with a Lanczos3 but I look forward to this dynamic implementation.

  • I have measured the minimum requirements for the card using GPUZ running in the background. A GT 240 DDR3 1gb is all you need, and will be running at~40 percent capacity. I recommend the following: Intel i7 hyperthreaded (4 coresx2=8 cores) 16gb ram (ram is cheap) and the gt 240 1gb DDR3, I have a GT 240 1gb DDR 5, and the DDR5 is pretty zippy, but I don't see any real world advantage in premiere pro. If you see a deal on the DDR5, grab it, otherwise the DDR3 is fine. Anyway, for $30-$35 you can't go wrong.

  • One nice thing about looks is that you can apply multiple modules into a single render pass so that you can avoid clipping highlights in one module if a later module brings them back. (There's also auto-shoulder which I put at the end of almost every render chain.) You can even use multiple Colorista II modules in one chain. (For example, I sometimes apply Colorista II to correct skin tones so that I apply Cosmo as a sort of secondary CC and then do my main color correction pass in a later Colorista II module. About the only think I wish Looks would do that it doesn't is power windows. I don't know why it can't implement these, since the tools are all there. Maybe Looks 3.

  • @liquidify - you are on-the-money in the workflow assessment, but I cannot comment as to the quality differences in codecs. Keep in mind that DSLR PPU does not do any conversion itself - it is actually ffmbc that provides that service. The codecs are all similar in nature, quality and format but conversion times vary, etc. Cineform is a unique codec in as far as its handling of metadata, but I do not own that codec so cannot comment on speed, handling or quality. I personally believe that all of the pro codecs provide similar results - and certainly all are sufficient for the footage we are working with!

    As far as RAID goes, yes, RAID6 is superior, but most casual users won't opt for larger arrays where RAID6 is possible. I prefer RAID10 for long term online storage and write to tape (lto-4/5) for archiving. My point is really that you just want to stash a copy safely somewhere for posterity, and work with your live files on a fast RAID0 array if possible. You just happen to be a lot more in tune with your storage requirements than most ; )

  • @DrDave Where can one find this "hidden feature, the superscalar resizing".

    Are you saying that enabling MPE hardware mode changes the premiere resize tool to use this better algorhitm? Or does one need to do something else to enable it as per the "hidden" in the name?

  • Yes, I found that CUDA scaling looks better than the software version.

  • You need to check a couple of boxes like maximum render and maximim bit depth. I can't recall which one turns it on, since I always use both. There used to be a bug in Cineform with scaling. I tested it myself, and found that unless you rendered out the video, THEN applied scaling, motion,etc, there was a big drop off in quality. Don't know if that fixed that, since I use the CUDA now. Just drop it on the timeline and play.

  • @JRD - Unless you have some good disc arrays or snappy SSD drives (and maybe video accelerators to help with the load), working with footage averaging 140-150Mb/s bandwidth and converted to 4:2:2 color space can quickly stress the limits of what disc I/O can support (and sustain in realtime). Any extra overhead caused by effects, coloring, transitions, etc will only add to the burden of realtime processing and your editing session can suffer greatly. Without some smokin' fast discs lengthy and involved editing sessions with high-bitrate footage can become quite painful.

    If you are fortunate enough to have a fast disc array, an AJA, Matrox or Blackmagic card and a beefy workstation then you probably won't feel the rub until VK hacks the GH2 to do 4k@120FPS ; )

  • I have a fairly new PC (GTX580 card and 6-core 3930K CPU), and tried out Cineform Neoscene last night. In Vegas pro 11 with a test clip (shot on Cake 2.0) using curves, MB looks and Neat NR, it took about 3 times as long to render the Cineform clip than working with the native GH2 file. So for me, Cineform wasn't very useful.

    I did try some extreme grading to see if Cineform looked better. Colour seemed less blocky but everything looked much softer. It's possible I'm doing something wrong, but I don't think Cineform is for me.

    If you have to have a format that works between say Vegas and AE, why not windows .avi? Apart from the humungous file sizes of course... but that's the only format that seems to play happily between these two applications, and Vegas really flies if I use .avi files rather than native mts. I'm not an expert on colour spaces etc, so this is probably a really stupid question, but

    1) Is windows .avi a good choice for intermediate rendering / going between Vegas and AE and
    2) In Vegas, why is windows .avi much quicker to edit than Cineform?!

  • @Macalincag how do you convert to cineform with AME? I can´t see the cineform presets after I install GoPro Free application? Thx!

  • @MrEdd

    For me, I installed it and they show up as available options in AME. One for Cineform AVI, and one for Cineform MOV. Strange thing is the Cineform AVI option has more "options" (codec configurable settings) than the Cineform MOV option.

  • Is there any more insight on whether the GoPro free cineform is inferior to the old Neoscene cineform codec?

  • @dbp I would not assume that the codec itself is in any way inferior until someone shows otherwise. So far I have not noticed a difference between the files saved with one vs the other and I go back and forth on different systems all the time. I was not looking for differences, though.

  • @Mark_the_Harp Cineform is available in two container formats: .MOV and .AVI. When you talk about using "AVI instead" which other codec are you using in the AVI container?

    In my experience, if I have Cineform files saved in a .AVI container, they perform better than .MTS files at the same bitrate if I checked "I-frame only" box before encoding. Were the files you compared at different bitrates? Or did you use a .MOV container for the Cineform files? Either of these things could be a factor.

    As another reference point, yesterday I had a session mixing GH2 and older HV30 .M2T files. Some of these were converted to Cineform. Scurbbing was much faster for the Cineform ones than the 25 mbps M2T files, even on an older USB 2.0 external drive.

  • @Macalincag ty, wandering if AME 64x is the cause of the problem. Ill try to install GoPro once again

  • @dbp Um.....I've used the pro version of Cineform before and now I am using the GoPro version. I'm not sure why some are saying that it is inferior because it has ALL of the same settings as the professional version EXCEPT Uncompressed 4:4:4. So, right now I use Filmscan 2 with the free version (the same level as I've used with the pro version). I believe most people don't even go as high as Filmscan versions because I've always been told that it was overkill. If that is overkill...then uncomressed 4:4:4 is.... well ....you get the point.

    Anyways, motion wavelet codecs like Cineform or JPEG2000 compared to other types of codecs provide a lot less file sizes under the same quality at normal bitrates. They give your footage a more "organic" look IMO. I wish every camera manufacturer went with wavelets instead of what they currently offer. I think it's amazing we can now get this for free from Cineform. Don't let anyone fool you....the GoPro is the very same codec as the other versions they've put out. I (and most of you) don't need the 4:4:4 Uncompressed version. There would be no benefit. jm2c

    EDIT: BTW, I use TMPGEnc to batch process all of my footage. It is very, very fast.

  • Ian, the early release (very first one) of GoPro cineform Studio had a codec version or constraint that limited the export bitrate to 3x the original file. Somewhere on David Newman's cineform blog (old one) or one of his many forum posts there is information explaining they had issues with that and decided to introduce the unrestrained codec in the post versions of GoPro Studio, which was also very buggy in the first release. So yes, now you have a free cineform codec which others have paid $129 for in Neoscene + a recent extra $49 upgrade. The only benefit of that is the ability to convert using only i frames.Thank you David, hope you enjoyed the "Gopro cash buyout and shares" i guess there will be no refunds forthcoming to the Neoscene purchasers.