Personal View site logo
What Bitrate level of DNxHD conversion?
  • dancerchrisdancerchris

    I am tired of all the editor problems dealing directly with our beloved GH2 .mts files. I want to convert to intermediate like DNxHD and would like to know what Bitrate should be adequate for capturing all the detail that a good Hack (in my case Moon T5) has. Any suggestions. Is it the same amount of the max bitrate in the clip? The DNxHD files are quite a bit larger, so minimizing this value will help to save space.

    Thanks,

    c

  • 9 Replies sorted by
  • mrbillmrbill

    We use dnxhd120 for most dslr imports

  • MrZzMrZz

    @mrbill @dancerchris I would recommend DNxHD too. Use 120 for most cases and if space is a factor. If space is no constrain, use DNxHD 180 to keep every bit of quality you can and if grading is a factor in the workflow later on. Both Bitrates edit very well. XMedia Recode (free program) does avery good job at transcoding this, batch conversion, Audio conversion etc. http://www.xmedia-recode.de/en/index.html

  • mrbillmrbill

    btw, dnxd120 is 8-bit. the 10bit flavours have an x designation at the end - eg dnxhd185x.

  • MrZzMrZz

    @mrbill correct! but I think in the case of GH2 (which is 8-bit) that would be overkill ;)

  • mrbillmrbill

    I agree - except in chromakey instances

  • rNeilrNeil

    @MrZz ... even with an 8-bit image, there are a lot of times where the processing of it may be complex that using 10-bit for the processing "stream" is quite useful. The vastly raised precision of the intermediate computations allows for a final result that is at least somewhat better than you might have had otherwise.

    I've been a stills portrait pro for 38+ years. When digital first came in, the only thing we could shoot was 8-bit jpeg. It was easy to test that for simple changes, there was no difference, but heavy duty work in Photoshop definitely benefited from taking the time to change to 10-bit. The resultant images, back in 8-bit, had smoother transitions that were VERY obvious when you had enlargements to make.

    Now ... using 10-bit processing on an 8-bit jpeg doesn't compare at all with say the 14-bit RAW file of a D3, quite true. But with most mid/lower price video cams we are essentially still shooting 8-bit jpegs. Which is why the Adobe apps pretty much assume doing all computations in 32-bit floating if they can. You can't make a silk purse of a sow's ear is the old expression ... but with higher bit intermediated computations you do get the best out of that sow's ear ...

    Neil

  • MrZzMrZz

    @rNeil @mrbill I won't argue with that! I think if it's clear that you will go through a heavy online postproduction workflow (Heavy grading, Animation VFx etc.) and disk space is no problem, then you should definitely convert the footage to higher bit-depth aswell as with better quantization (most people use 5dtoRGB for that, converting to, say ProRes 444).

    @dancerchris as you probably know, DNxHD is a lossy codec and since you stated that disk size is a factor to consider, I still think 120 will suffice. It's definitely better to go for 185X if you wan't to do some grading but it's quite a bit more disk-space and probably better to go for a lossless codec in that case or just choose a offline editing workflow, using for example DNxHD 36 and then relinking o the original footage later on.

    Oh, in what framerate did you actually shoot? because that's one of the factors which affect the available DNxHD bitrates. e.g 25p --> DNxHD 120/185/X 24p --> 115/175/X

    I think in the end it always comes down to the hardware constraints you are facing and what you want to do with your footage, correct me if I'm wrong guys.

    Cheers!

  • rNeilrNeil

    Yep. Tradeoffs abounding ... which is to say, Life. ;-)

    Neil

  • balazerbalazer

    There are benefits to working in high precision, but converting 8-bit h.264 to 10-bit DNxHD has nothing to do with that and provides no benifits, assuming that levels are preserved in your conversion. (ffmpeg for example preserves levels) You may wish to use 10 bits for intermediate files if there are color correction or level altering stages both before and after.

Add a Comment