Using Intermediates In Post Production
The use of intermediates is a well known practice for major Hollywood motion pictures post production so the question is not if you should be using them, but why aren’t you using them?
I’ve tested the two options known on the PC side of things – Cineform and AVID DNxHD intermediates and I can tell you, using intermediates is a far better and more efficient way in the long run to edit if you want increased performance, real time playback and more accurate color grading capabilities. Those using Final Cut Pro are doing the same thing by converting their DSLR clips to Apple’s ProRes.
I initially started out using Cineform NeoScene, but after extensive testing I just can’t recommend it until they resolve a glaring bug in their capture utility. Here’s the results of my findings: Currently HDLink removes the beginning and end of clips thereby removing frames if comparing to original clips. For long form post production work that is a serious issue. Instead, I’ve gone to using AVID’s DNxHD codec and transcoding my highly compressed native GOP h264 DSLR clips with MPEG Streamclip to the frame based DNxHD intermediate. For those unfamiliar with MPEG Streamclip, it’s a free utility that utilizes Quicktime to do what it does. And it does a fantastic job in the process of encoding to AVID DNxHD (or any other codec Quicktime can access).
According to AVID: “Native HD camera compression formats are efficient, but simply aren’t engineered to maintain quality during complex post production effects processing. Uncompressed HD delivers superior image quality, but data rates and file sizes can stop a workflow dead in its tracks. Avid DNxHD (intermediates) delivers both efficiency and quality without compromises.”
There are compelling reasons to use the AVID codec over Cineform – ie; it’s truly cross platform. Any NLE that can read quicktime .MOV files along with installing the free codec can access and play them back. Which means ANY Windows or MAC based NLE can be used to edit the clips if it can read Quicktime MOV files. With Cineform, you pay for their codec and utility and have to jump through their activation hoops – which are known to not work as well as advertised.
Presently, I work in Premiere Pro CS5, and previously in Vegas Pro 10 32 & 64 bit and along with the latest version of Quicktime player – the clips work wonderfully. I can play back 7 streams of DNxHD 720p 110mbps 10 bit clips on the timeline stacked in Best Full preview playback in real time (with no effects applied). That speaks volumes to me at the quality of the codec. Bumping up to 1080p strains my Q9400 Quad core with 8GB RAM with a Raid 0, but it does playback well enough to get things done. I encode to 720p 110mbps 10bit to give me as much room as possible to grade my clips without degradation to the clips. It’s the same principle that Cineform uses when transcoding to their AVI based intermediate.
There are those who like Cineform, and I’m sure it’s a good product. I even had a license for NeoScene for over a year, but after I found the glaring bug that was first denied publically by those involved at Cineform, yet privately, was told by tech support that the bug was reproducable, I felt it was time to find another option. And the slap in the face? Cineform Tech Support’s response as a workaround was to edit the native highly compressed files. That sealed their fate in my work. Hence my using AVID DNxHD intermediates as the foundation for all my post production.
Here’s my workflow procedure to encode to AVID DNxHD from original clips. You need to have the following installed in this order: Quicktime Player (The latest version), AVID DNxHD codec (Latest) and MPEG Streamclip. Just as with Cineforms HDLink utility, you can batch encode your clips with MPEG Stream Clip and save them to another folder.
- Once my original native footage are copied over, I open MPEG Stream Clip, select List and then batch list from the menu at the top.
- Select Add Files at the bottom, and navigate to the folder with the native DSLR clips. Add all the files to encode.
- Then choose “Export To Quicktime” – do not check any of the boxes presented. Uncheck them if any are checked.
- At that point you will be asked to select the folder where you want the rendered intermediates to go. Once you select that folder, a window will open that shows your encode options. It will probably default to “Apple Motion JPEG A”. Drop that box down and select “AVID DNxHD Codec”. If it doesn’t show as an option, you need to make sure you have installed the DNxHD codec. If it shows, slide the Quality Slider all the way to the right (100%). Select the options button. You should leave the RGB radio button as the default, Alpha is None, and then comes the place you select the bitrate and color space (8 or 10 bit).
- My workflow has been to select the 720p 110mb 10 bit 29.97 or 720p 90mb 10 bit 23.97 option when delivering my clips to the web. I choose this option to still give me HD quality footage and to give me a slighly smaller clip in pixel aspect ratio (PAR). Whatever option you choose, click ok.
- Next, I select the Frame size (in my case, 1280×720 radio button) in the Frame Size area. Since my footage is being resized, I select Better Downscaling under Frame Rate.
- For audio, I select Uncompressed Stereo and 48khz (DSLR cameras usually shoot in 44khz, but I select manually just to match to the project settings I use – I will then sync in post with my Zoom H1 audio recorder as needed which is also recorded at 48khz 16 bit using Singular Software’s PluralEyes). With regards to Stereo, some shooters who only use one chnnel when recording audio with their video may want to change that to mono. It does make things easier once in post not having to try and combine audio channels after the fact.
- Once all that is done, Select the “To Batch” button. You will then see all your clips added to the batch window to be processed according to your settings. If you’re sure of your settings, click on the “Go” button and this will start your batch render. Go make some coffee, tea or what ever. Depending on the number of clips, it’s going to take awhile. I typically setup an overnight batch encode when I have a number of clips to convert.
Depending on the number of CPU cores you have, you can actually have more than one clip encode at a time. It does speed up the process somewhat.
Once the original clips are encoded to DNxHD, I choose to archive the original DSLR files to blu-ray discs once the project is finished along with the project files and all other associated files – this allows for pulling the project to edit again if needed since originals are also in a .MOV file wrapper – just like the DNxHD intermediates. This will free up hard drive space needed while in the actual editing environment. When transcoding clips from Canon’s 5D, 7D, and T2i, it works wonderfully and renders a little faster due to the clips already being MOV wrapped h264 files.
So you may be asking why go through all this in the first place? – Read AVID’s explanation here for the details.
I have personally endured editing compressed footage and it doesn’t hold up well for the most part in color grading. It also taxes the CPU extensively – even the latest i7 CPU’s. Those who have slightly older hardware will benefit greatly from going through this process of getting your footage encoded to a more friendly, less CPU taxing intermediate file format. In addition, if there’s any collaboration to be done in post, the AVID DNxHD clips can be brought into ANY NLE. Along with an exported AAF or EDL, you can hand off your project to another editor using a different NLE.
This is how Hollywood edit’s multimillion dollar blockbusters. You too can use the same techniques to edit your projects with alot less frustration.
I highly recommend using intermediates when in post production. It just makes sense.