With the explosion of (mostly Canon’s) video HD dSLRs in the last few months, purchase decisions for video editors have shifted, depending on which editor can deal with h.264 the best way. Until recently, users had to either use “proxy” files, or transcode to an intermediate format. Then, Adobe’s Premiere Pro CS5 came in to change this by being the first video editor to fully utilize nVidia’s CUDA technology and achieve real-time playback for Quicktime dSLR, and AVCHD footage.
This is by far the main new feature on the video side of CS5. With a supported nVidia GPU you can easily get real-time support not only for “naked” h.264 footage, but also when using some effects. To get advantage of the full acceleration you need an nVidia card with 768+ MB of VRAM. Non-supported nVidia cards that do have as much RAM can be supported via a hack. We asked Adobe why such a high requirement of VRAM, when decoders like CoreAVC don’t require as much, and this is their response:
“There’s nothing else on the planet using CUDA the way we are. CoreAVC uses the GPU for one task – decoding a particular video format. We are using the GPU for handling real-time effect filters on multiple streams of HD, 2k, or even 4k footage. This isn’t just one task – there are over 30 different real-time effects. Doing this requires more RAM.”
The good news don’t stop there. The re-written h.264 playback engine (now called “Mercury”) is very fast in “software” (DVXA?) mode too. While with CS4 we’d get about 10 fps with 1080p dSLR footage on our Quad Core PC, with CS5 we get full speed in “draft” mode, and about 16 fps in full 1080p mode. These are speeds that my own video editor of choice, Sony Vegas, can only dream of right now. Stability-wise the Quicktime formats are more stable on Premiere Pro than on Vegas, however, I did manage to crash the whole Vista PC when Premiere Pro froze while switching from the “Editing” to “Color Correction” screen mode.
Other new features on Premiere Pro CS5 include native support for AVCCAM, DPX, XDCAM HD 50, and AVC-Intra. The RED format has seen an update too. dSRL resolutions/frame rates are now supported out of the box in the project settings. Professionals now can read/save project formats for FCP and Avid MC. This is a key feature for studios that must send whole projects to colorists, or effects and graphics artists, who happen to use different software than CS5. Also, Premiere Pro CS5 comes with support for Adobe Story’s and Adobe OnLocation’s metadata, making it easier putting together a first rough cut.
Finally, I found a bug with the h.264 encoder on CS5. If you check the “PSP” MP4 container type (instead of the “iPod” one), the created h.264 video won’t play with any application I tried it with (I tried CoreAVC, Quicktime, and VLC). CS4’s “PSP”-type MP4 h.264 files used to work just fine. I generally prefer the PSP type because I always export for my Sony PS3, which also prefers that type of MP4 h.264.
Premiere Pro CS5 loading a clip from my upcoming project, a music video for the band “Hello Monster”
On the After Effects CS5 side, the biggest new feature is its 64bit support. Along the welcoming previewing speed abilities because of 64bit, this also means that your old 32bit plugins won’t work anymore. Thankfully, many plugin providers have already updated their plugins to 64bit, but many free or cheaper ones will probably never work again.
The included third party applications/plugins Mocha & Mocha Shape, and Color Finesse have been updated to versions 2 and 3 respectively. I personally did not like much Color Finesse v2 (it was crashy in addition to its few abilities), but v3 seems to be going to the right direction, with added support for LUTs. The Digieffects FreeForm companion app is new to the bundle: it allows for a layer to be manipulated in 3D space.
The Roto Brush is a very impressive tool: you create a very broad/generic stroke to tell AE where the foreground and background is. The app then can create a matte with an alpha, automatically. It’s not a perfect tool, and it has trouble with some blurry/fast sequences, but the fact that it generally works it already sounds like an engineering marvel to me.
Overall, CS5 was a very strong release for the video side of things. More so than the usual favorite of each major Adobe release, that is Photoshop. If only Adobe pays some attention to the kinds of video formats Photoshop is able to export at (it still can’t use all Windows-wide installed encoders, e.g. Cineform), and the usability of After Effects (when using too many layers) can be daunting. Premiere Pro is not super in ease of use either. In fact, user experience is the only reason why I still use Sony Vegas. If Adobe tackles that issue in CS6, then they’ll have an even bigger winner.
FTC Disclosure. Per the new FTC guidelines regarding bloggers and disclosure: Adobe sent me the CS5 suite about a month ago, free of charge, per a request I made to them for a review unit. I report my findings to other users, and I’m calling them as I see them.
But is the export options flexibility and bugs that Premiere Pro (and Premiere Elements) suck at and Sony Vegas gets it right.
Have we got to the point wher. A nice dSLR like the Canon 550D could be used instead of an HD camcorder? That would be amazing to kill two birds with one stone
Unless you’re shooting weddings or keynotes where tape is better because it can shoot continuously, then the 550D/T2i is indeed the best camera in the market today, in terms of “money vs abilities”. For $800 you get amazing things. Things that even just 3 years ago you wouldn’t think of being able to have for that kind of money.
dSLRs do not substitute camcorders in many scenarios.
dSLRs typically shoot in mono and their electronic and sensors are not optimised for video. You will never get from a dSLR like the Canon 550D/T2i the sensitivity of a camcorder like the Panasonic HDC-SD 300 and similar.
This makes a huge difference when for instance shooting your kids indoor with sub-optimal light and they move…
Moreover camcorders are much smaller and easy to handle, again a big plus for family or holiday shootings.
So it is not just weddings and keynotes…
Shooting a small movie with a dSLR is ok though for casual use, when you do not have a better device at hand. And it is getting better and better, I agree.
Edited 2010-06-02 09:58 UTC
That’s plain wrong, the only advantages that dsrl have vs camcorder are:
big sensors with GREAT and never seen before in consumer market light sensitivity
Small DOF
many many expensive lenses ’till you are bankrupt
That’s the case for a single-shot scenario.
Beware that dSLRs can shot 3-4 fps in photo-mode.
When you go in movie-mode (24-30 fps) it is a whole different story.
Shoot your films with a dSLR and digital HD camcorder, and compare for yourself.
Edited 2010-06-02 12:23 UTC
See webepisode 2
http://www.zacuto.com/shootout
Acording to this teste the dsrl are better that film as light sensitivity is concern
I have a canon camcorder HF10 and a canon t2i and I can say that the t2i is far better in low light
You don’t get the full advantage of the big sensor however. You do get the shallow DOF advantage but not the sensitivity advantage.
They don’t read the whole sensor and put together data from all the pixels into a lower resolution, such as filming at 21 megapixels on the 5DmII and saving in 1080p. They only store the needed pixels AFAIK so that leaves out a hell of a lot of sensor area. For 1920×1080 that will mean like you’re only reading 1/10 of the sensor of the 5DmII. So yes, much bigger sensor but you lose 9/10 of the sensitivity advantage ..
It’s waay too much data to fetch those 21 megapixels @ 14 bits (not times three since the pixels aren’t RGB-pixels) / pixel @ 30 fps and then group them together…
It would be 1102 MB/s read from the sensor ..
The 7D with it’s two Digic4 processors can do 8 FPS @ 18 mpx @ 14 bit.
No it will have higher sensitivity? ..
Even more so for the 5DmII.
(Edit: Correct myself with the post above. For DSLR video not all pixels are read so for the 5DmkII reading 1920×1080 of the 21 million pixels would leave out 9/10 and hence the sensitivity won’t be as good as the sensor size would suggest. Atleast until they catch up to being able to read all that data at such frame-rates.)
Mono is bullshit, lots of them can have an external mic plugged in to.
DMC-GH2 will have global shutter.
DMC-GH1 got decent contrast-detect focus.
The new Sony HD cameras (not the NEX, the Alphas) will be able to use the optical viewfinder and phase-detect autofocus during filming.
Edited 2010-06-02 15:47 UTC
Not all dSLRs have a mic-in, take for instance the Nikon D5000. Moreover I do not find it very practical to go around with an external mic to connect to my dSLR for casual film shooting.
If it’s important for you you get one which record stereo or can connect a mic.
If it’s not important for you then you won’t complain …
Rather simply isn’t it?
I’ve seen a few awesome DSLR setups with rails, hoods, mic, lighting, .. Those people obviously care. If you don’t then don’t complain, get what works for you =P
Internal microphone is likely to pick up AF noise anyway.
It makes me feel sad to see Linux video editors falling behind even more… supporting Nvidia CUDA for acceleration of encoding or even just rendering would be fantastic. I can’t say I’m overwhelmed with the admission that these pro-level programs couldn’t deal with H.264 video files properly, though, but at least they’ve got it right now.
Linux video editors are not falling behind on the pro scene. Many studios use linux and custom apps to edit their movies (just search for linux + hollywood in google). And there is for example Piranha from IFX which is a linux suite for video editing used by big names like Dreamworks and Pixar.
http://www.imartis.com/
Let’s hope for a consumer version.
Edited 2010-06-02 11:57 UTC
This begs the question: what is the best distro to use for that kind of software? RHEL? I use Ubuntu. Should I switch to Fedora?
Some of that software looks absolutely amazing. I am surprised that they haven´t decided to offer a consumer version of some of their offerings.
I, for one, would love to see that and would gladly pay for it.
I don’t think you understand the workflow of film, they are NOT being edited on Linux almost EVERY film is edited on AVID on Mac or PC.
Most visual effects are done on Linux through Flame, NUKE, Shake, Maya ETC
and most animation is done on Linux with Maya, Soft Image
Finishing and editing are totally different.
But most Linux editors SUCK!
I prefer the “official” company site, be it only in terms of aesthetics ^^ http://ifxsoftware.com/
In the end, everyone looses when any company uses CUDA instead of OpenCL.
Congrats to V-Ray people that opted for OpenCL, thinking about the costumer and the market.
I like the in depth and unbiased opinions expressed here. Hope the finished product is as polished. Good Luck.