Last week I had the privilege of conducting a webniar with NewMediaWebinarrs.com called DSLR Filmmaking Post Workflows. That webinar is now available on-demand from New Media Webinars for $25. That gets you the entire 90 minute webinar in HD, an audio podcast where we answered some questions from those attending, several Canon 7D video files for your own workflow tests and some useful links. There were a lot more questions asked by attendees than we had time to answer so I jotted down some quick answers to most of them and have posted below.
Are there any advantages to Canon’s H.264 over Nikon’s Motion Jpeg?
While I’m not sure of the exact data rates that the Nikon cameras shoot, H264 is a more efficient codec that Motion JPEG and better for compressing overall image quality into a small file. I’ve seen very little Nikon DSLR video footage come through the edit suite so I can’t say for sure but there’s a reason that almost everyone is shooting the Canon DSLRs over the Nikon.
Are there other alternatives to importing clips as reliable as Cineform Neoscene?
Cineform’s NeoScene is a utility for transcoding DSLR or HDV footage into either ProRes or Cineform’s own codec. Other transcoding alternatives would be any of the tools dicussed in the webinar, Apple Compressor, Magic Bullet Grinder, MPEG Streamclip or the Canon EOS Movie Plugin-E1 Log and Transfer plug-in for Final Cut Pro. While ProRes is a popular (and built-in) codec in FCP you could use Cineform as your intermediate codec of choice as it’s a great codec and has a successful track record, especially in digital cinema. Cineform recently announced the support for Realtime RT in FCP as well. While I haven’t used it, Cineform also offers an interesting product in First Light which is part of some of their packages. While a more expensive option than the NeoScene utility something like NeoHD offers some advanced color grading options for realtime color correction that can updated with changes that are reflected in your edit. Check out Cineform’s website for details on their products.
I would like to know about combining the footage with red material and also audio sync issues.
Intercutting DSLR footage with RED is easy as long as the RED is converted into a matching format. If both formats are, for example, 23.98 ProRes 422 then it’s easy. The key to making them match would be in the final color grade and with some effort a DSLR camera can be a nice B-camera to a RED. If you want to match back to the RED R3D files in online then just put the RED shots on their own video layer so you can pull that out of the edit for online conform. Audio sync shouldn’t be much of an issue if you are matching frame rates. The key for the smooth mixing of footage in post production is to match frame rates.
I’d like to talk about artifacts due to the algorithms and h264 compression, how to avoid that for clean vfx work?
I don’t know if you really can avoid H264 artifacts in their entirey since you have to shoot H264 with the Canon DSLRs. Just be aware of what could cause artifacting while you shoot, (busy backgrounds, fast camera movements, extreme contrast) and try to avoid them. Then move the footage out of the 8-bit H264 world when you get to post.
What’s the best workflow for 24p in FCP?
To get 24p then shoot 24p. For the 5D that means be sure it’s updated with the 24p firmware. And when you shoot 24p you are really shooting 23.98 which is television friendly as opposed to true 24p which is only needed if you are going to make a film print in the end and gives you a 1:1 ratio between video frame and film frame. In that case it’s really the audio that you have to worry about so it is recorded at the proper rate to stay in sync. That would be a case of conferring with your audio pro. But really … how often are people actually going back to film these days?
Please talk about “undercranking”. And doing this in camera, versus speeding up footage in editor (FCP – apple + J shortcut)
You really can’t under crank any of the DSLRs that I’m aware of. That involves shooting at a slower base frame rate and then conforming that to your normal rate thereby speeding up the shot as opposed to “overcranking” which is shooting at a faster frame rate and then conforming that to your base rate. This overcranking is, in essence, what happens when you shoot 720 59.94 on the 7D and then conform that via Cinema Tools or Magic Bullet Grinder. There’s more frames in a second of footage so those 59.94 frames get stretched over time (in essense but really played back slower) to 23.98 which slows the footage down. Since you can’t “undercrank” the DSLR you have to speed the footage up in post.
What is the best way to archive DSLR media for later re-use?
The archiving question is one that is constantly asked with digital acquistion and has yet to answered. Of course multiple hard drives are the simplest and most affordable way but hard drives crash and aren’t good for long term storage. If used for long term storage they need to be power cycled and have data read across the drive on a regular basis (once? twice a year?) to help with reliability. Beyond on that Blu-ray might be an option and with the new 128 GB Blu-ray spec that’s a lot of data on a disc. Probably the most reliable thing is the old standby LTO tape.
Are there any cameras that record directly in Pro-res?
Arri’s new Alexa can record directly to ProRes in addition to its RAW format. I don’t know of any others but the AJA Ki Pro is an external unit that the output of a camera can be plugging into and then that unit will record ProRes but from my understanding the Canon DSLRs don’t output a good, clean HDMI signal then that might not be the best option.
What is the main advantage of using DSLR vs Video Digital Cameras?
A big imaging sensor for shallow depth of field without using a cumbersome 35mm lens adapter would be advantage # 1. Those big sensors also produce an amazing video image for the price. They are also more inconspicuous than a pro-video camera and can shoot in places a pro-cam cannot. And the images make many want to lick the screen. There are also a lot of disadvantage but since that wasn’t asked in the question we’ll leave that for another discussion.
When you transcode footage from 8 bit to 10 bit (Pro Res) are you getting true 10 bit since you started with less color info?
ProRes is a true 10bit codec so you are “uping” the image out of 8bit space and any work you do on the image from that point forward is working in 10bit color space and can take advantage of what 10bit can offer. But for a balanced perspective on why staying 8bit in the right application might not be a bad thing read this article. Isn’t all this confusion wonderful?
How does it compare to the EOS7D which is being recommended (kind of) here. The other Webinar slanted towards Canon 5 Mark II I think.
Most everything discussed in the webinar applies to all the Canon DSLRs since the files that come out of the camera are virtually identical. The 7D has an APS-C sensor which is smaller than the 5D’s full frame sensor. A smaller sensor means less shallow depth of field overall (but that’s so dependent on the lens and aperture), lesser low light performance as well as more noise. Plus when you are using full frame lenses on the 7D’s cropped sensor the focal length of the lens will be different. But the 7D is more affordable (as is the 550D/T2i even moreso) and a great camera.
Is it feasible to put backup files on DVD or does this become to cumbersome?
IMHO DVDs are so limited in storage capacity and slow to burn it would take a very long time to backup a big shoot. You could use something like Toast to automate how it burned a full shoot to DVD and let that handle how clips would span multiple discs but I would just buy an extra hard drive for backup keeping in mind that hard drives need to be powered up and have their data read to keep them operational over the long term. I have some backup drives I power up a couple of times per year.
How does ProRes have an Alpha channel?
As of this writing, only ProRes 4444 has an alpha channel. That’s what the extra 4 is for in name. Though keep in mind ProRes 4444 is overkill for DSLR work. You could use ProRes 4444 for individual effects shots that need a lot of post processing or to carry and alpha around but overall it’s overkill.
What is the best codec to transcode to if you are editing for the big screen? would ProRes LT still be good enough?
I personally would move above LT for the big screen to regular ProRes or ProRes HQ. While the added data rate of HQ is rarely worth it for DSLR footage this might be a case where it could help the files hold up. That said I have heard of ProRes LT being projected in the theatre. That’s a question that’s ripe for testing! Personally I would be very comfortable with ProRes HQ knowing that I had that little extra data in the files and knowing that image was going to be projected in the theatre in the end even though there might be no discernable advantage to ProRes HQ over regular ProRes. One advantage would be in my head as I was editing as I wouldn’t be wondering if ProRes HQ would be just marginally better image when projected in the theatre. And if the editor is happy in their head then everyone is happy in their head …. right?
When filming a wedding 1920×1080 24p which ProRes would you recommend using and why?
By 24p I assume you mean 23.98? For a wedding I would be more than fine with ProRes LT. Great image quality vs. file size with LT and a lot of weddings end up on DVD or the web.
How does Grinder compare to the Canon movie plug-in?
Magic Bullet Grinder allows the assigning of your own timecode to all of the clips in a batch while the EOS E1 plug-in uses the timestamp to assign each clips its own timecode. EOS E1 wins in that respect. Grinder allows the simultaneous transcode of a low resolution clip for offline editing. Grinder will also conform 720 59.94 clips to your editing frame rate (and uprez to 1080) for slow motion. And Grinder is incredibly easy to use so someone with only limited knowledge could set up a batch. I looked at the tool in-depth here.
To make things happen quickly, can you shoot flat and then use Magic bullet looks presets rather than making micro tweeks on each frame?
Sure, you can Magic Bullet Looks any shot, regardless of it being a flattened shot or not. But remember, a preset is just that; a canned look that everyone has access to. The beauty of color grading is that you can tailor a look to your specific need and creative sensibility. Plus you can balance the overall look of an edit and better match the look from shot to shot. Presets are the lazy man’s way of color correction.
Is there ever an advantage to transcoding in Compressor and then importing to FCP instead of transcoding within FCP?
Probably not any more than the fact that Compressor can run independently of FCP and take advantage of multiple cores via a QuickCluster. But the Canon EOS E1 plugin adds good metadata like a reel number and timecode so I prefer to use it over Compressor. An in-depth analysis of image quality differences between some of these transcoding applications hasn’t been done as of yet but just a casual observation doesn’t show much difference to my eye.
Is there a way to fix anti aliasing/line skipping in post? fine textures in clothes, eye brows, hair, wood, meat (texture) is things i have seen in commercial i been editing. pixels a flickering in these fine texture areas?
That’s a good question. A subtle blur can help with some of the moir© patterns that might occur . You could use a region blur to position the blur over select parts of the frame. Often just a very slight blur can reduce an annoying flicker.
If you pull files in via Log & Transfer, does it hold the timecode metadata?
Yes, the timecode the Canon EOS E1 plugin assigns to the clips are based off of the time stamp the camera assigns to the clips when shooting so make sure the cameras are set with the correct time of day.
Any quality difference between transcoding in FCP (during ingest) and Grinder and Compressor?
Not to my eye but AFAIK there hasn’t been an extensive comparison posted on the web comparing the various transcoding methods.
Can Grinder conform to 25 fps?
As of the first release it cannot.
How to do a multicam(music video) edit in FCP using dslr material?
Once footage is converted into your editing format then DSLR material is just like any other piece of media so no special care has to be given to the clips at that point. If you’re asking how to do a multicam music video edit in general then that’s way beyond the scope of this post. But you should look to auxiliary timecode for one (here’s an article I wrote about just that) or buy Mitch Jacobson’s Mastering Multicamera Techniques book which, full disclosure, I had a very small hand in helping him put together.
Any problems with Grinder? I’ve experienced altered audio. Sounds like the Chipmonks.
I have seen no problems with Grinder altering audio.
Could one make batch convertions in MPEGStreamClip?
Yes, MPEG Streamclip does have batching capabilities. Look under the menu List > Batch List and you can then use a saved preset for the batch.
I have a question, are there any tips for syncing audio from a sound recordist – is it best to sync the rushes first or to add sound after the edit?
I would always sync the sound first. My techniques involve making a sequence with synced sound and then either copy/pasting from one timeline to the other or loading the synced sound sequence into the Viewer/Source monitor and using that as a source clip (careful NOT to nest the edit in FCP). I don’t use merged clips in FCP as they seem to always cause problems if you have to media manage and/or move an edit. The new version of PluralEyes will also create a new master clip that uses the recorded production audio. While I haven’t tried this feature yet it’s a brilliant idea.
Can you do the frame rate conversions in bulk?
Yes, by using the Batch Conform command in Cinema Tools but BEWARE when you choose a single clip in a folder full of clips it will conform EVERY clip in that folder as that’s how Cinema Tools rolls. And it’s a destructive operation, changing the file itself. That’s often okay as you want to slomo all of your 59.94 clips but it’s very important to be aware when you perform a destructive operation to your media.
What is the best way to make slow mo for 5d, shot on 24p?
You can apply speed effects or motion effects via Media Composer or FCP’s built-in speed tools. But fake, post-production slowmo often looks a bit fake, stuttering or showing some nasty artifacting. That’s just part of slomoing in post. I heard Apple Shake has some nice technology built in that does a decent job and that tech might be in Motion as well. Media Composer has several different rendering methods for slomo that can be experimented with.
How do you deal with mixing 720 and 1080 footage?
Your editing application will usually either scale up or scale down your footage depending on what your base editing format is. In FCP that’s your sequence where in Avid that’s your project. And they do a pretty good job of up or down rezzing footage to match.
Why not just Slo_Mo in FCP or Motion?
When you shoot 59.94 and conform to a 23.98 frame rate you are getting a frame by frame “slow down” of the clip which means it’s a clean clip without the artifacting or frame blending that happens when you try to slow down footage in post. If you conform a clip via the Cinema Tools method and then take that same clip and slow it down in FCP and then step through the image frame by frame you can see this happening.
What is the difference between using cinema tools or grinder to do slo-mo versus using the speed tools in the motion tab of FCP?
See question 31 above. Cinema Tools or Grinder is actually CONFORMING the clip to a new frame rate and giving you a new clip slowed down to that frame rate vs. just an effect where the software has to interpolate what to do with in between frames or figure out how to create new frames and how to handle the changes from frame to frame.
how to deal with mixed footage of a Canon 5D together with say, Sony HVR-Z1 (=50i)?
I think the most important thing when mixing any other camera with a DSLR is to match the frame rate … and usually that means 23.98. A traditional “video” camera shooting a native 23.98 frame rate is going to have more of a “film look” (to use a cliched term) and be easier to match. Then it’s often a matter of a good color correction to match the footage. That’s where a trained colorist can work wonders.
4444 too high for DSLR?
Yes, ProRes 4444 is overkill for H264 originating Canon footage as this codec was designed for higher end, full color space workflows (It’s 4:4:4 as opposed to 4:2:2) and to carry an alpha channel. If you’re passing shots back and forth to vfx then it might be worth it but for general DSLR usage it’s overkill.
How do you obtain a flattened image?
Check this FreshDV post that links to a number of articles about setting up your DSLR for flattened shooting.
If using native editing ? whats you color grading workflow?
The only time I use native editing is in Avid Media Composer (and that’s rarely as the native H264 performance isn’t perfect) or Adobe Premiere Pro CS5. Then I just use the built-in color correction tools. In FCP I’ve already transcoded my DSLR footage to a flavor of ProRes so then I just send the edit over to Color or use Magic Bullet Colorista from within FCP.
What specs are required for smooth native playback?
True, smooth native playback of Canon’s out-of-the-camera H264 files require Adobe Premiere Pro CS5 and it’s Mercury Playback engine for “smooth native playback.” Otherwise it’s transcoding into ProRes for a Final Cut Pro edit or DNxHD for an Avid Media Composer edit. While Media Composer 5.0 does support native H264 playback via its great AMA architecture I’ve found less than smooth playback on the slower of my two Mac Pro’s so I usually transcode the material once in Avid and I’ve made my selects.
What hard drive do you recommend for a MacPro for native editing?
None since I wouldn’t recommend native DSLR editing unless I was working in Premiere Pro CS5. As for intermediate codecs like ProRes or DNxHD, they can perform well with any of the hard drives that are geared toward video editors, like the units from G-Technology or CalDigitg the. Firewire 800 can work but the best bang for the buck is to invest in an e-sata card for you Mac Pro as many of these drives now come with an e-sata connection. A RAID configuration is faster than a single drive but there are different types of RAID that offer speed vs. redundancy though that is way beyond the scope of this document. A couple of internal hard drives in the MacPro can be raided as well for good performance.
I hope these answer most of the questions left from the DSLR Filmmaking Post Workflows webinar. If you didn’t see your question it was probably answered in the podcast that was recorded after the webinar. The full package is available for $25 from New Media Webinars. Thanks to those who attended. I jotted down these answers over early coffee so if I’ve missed any vital information then feel free to follow up in the comments below.