DaVinci Resolve

Blackmagic Design has always been on the cutting edge of technology with an eye towards shaking up the status quo. The acquisition of DaVinci, with the port of Resolve to the Mac platform, promises to be just that. Resolve 7.0 is available for both Mac and Linux and comes in three flavors (not including the workstation itself): Mac software ($995), Mac with a custom-designed Blackmagic Design colorist’s panel ($29,995) and a Linux version with the panel ($29,995 plus $19,995 software). Linux pricing is in addition to the necessary hardware and integration costs supplied by authorized DaVinci/Linux resellers.  The Mac version has sparked the most interest, but even the Linux configurations represent a significantly reduced system price from years past.

DaVinci Resolve offloads processing to a single GPU on the Mac, but can be used with multiple GPUs in a Linux configuration. Resolve achieves its real-time performance on the Mac thanks to leveraging the CUDA technology of the NVIDIA cards. Blackmagic Design recommends installing two NVIDIA display cards (a GT120 for the interface and monitors and a separate Quadro FX4800 for accelerated processing). Currently only Blackmagic Design video I/O cards will work, so if you install two graphics cards, a Decklink card and an external storage controller card, you will have maxed out the available slots on a Mac Pro. The FX4800 card is a double-width card and the recommended Decklink HD Extreme 3D Plus comes with a second bracket for the HDMI connectors, so not everything will fit. You’ll probably have to forgo the HDMI connection for the Decklink. On the bright side, the FX4800 has been replaced by a single-width Quadro 4000 card. So, juggle your Mac Pro slots wisely. (Download the Resolve manual and Mac configuration guide here.)

Installation

Resolve is a painless installation on most machines, but I did  have a minor conflict on one system tied into a shared storage network. The networking components of Resolve conflicted with Command Soft FibreJet; however, a friend has Resolve running on a Terrablock SAN without any issues. In addition, the installer adds a number of Blackmagic Design utilities, QuickTime codecs and Final Cut Pro Easy Set-ups. It will flash any present Decklink card with a firmware update and prompt you to install updated drivers for existing NVIDIA cards. Despite my network issue, most of these are harmless; but, to be on the safe side, clone your hard drive before starting. Since new FCP Easy Set-ups have been added, your FCP Preference files will also be deleted and will need to be rebuilt the next time you launch Final Cut Pro. Resolve uses a USB license key (dongle), so no internet activation is required to get started.

Despite what sounds like very rigid configuration guidelines, DaVinci Resolve actually runs on a wide range of hardware, including laptops. The one absolute is that you use a monitor with a resolution of 1920×1080 or 1920×1200. The interface doesn’t scale, so a lower display resolution, like that of a 20” Apple Cinema Display will cause the interface to be cropped and unusable. I tested the software on an 8-core Mac Pro with only the single, stock NVIDIA GT120 card. The trade-off in not having the second NVIDIA card is a loss of real-time performance, but it does work relatively well. Likewise, ATI cards should also work, but according to Blackmagic Design, this is currently untested and considered an unsupported configuration.

Getting to know Resolve

DaVinci Resolve is designed as a full-fledged, standalone color grading application. It is not a plug-in to a host NLE or built around a roundtrip workflow. In addition to the color correction tools, there are ingest, conform, render and output modules. These interact with the Decklink card to accommodate both a file-based and VTR-based workflow, even if no NLE is installed on that workstation. Sequence information can be imported from EDLs or AAF files, but an optional Avid component ($500) must be added to permit access to Avid DNxHD media. Only a single layer of video is supported on a DaVinci timeline.

The basic color grading concepts are not that different from other systems. The adjustments can be applied to both primary and secondary levels, which are handled as nodes. More on that in a minute. Unlike most other products that display lift/gamma/gain correction as a set of graphical color wheels, Resolve uses sliders for luma/red/green/blue values for each of the lift/gamma/gain settings. In addition, there is a separate luma control section, as well as curves. If you intend to run Resolve with only a mouse or tablet, then you will be in for frustration. The l/r/g/b sliders do not gang or work in a push-pull manner when using a mouse, so if that’s you, then curves will be the section you’ll use the most.

The only third party control panel that works is the Tangent Devices Wave ($1,495) and I feel that’s the absolute minimum surface to buy, if you intend to use Resolve frequently and productively. Of course, if you intend to establish a dedicated DaVinci suite, then it makes sense to spring for the more expensive Blackmagic Design control surface. Another consideration for a real grading suite would be scopes. Resolve includes a set of integrated software scopes (waveform, parade, vectorscope), but for better signal monitoring, I would highly recommend equipping the room with a Blackmagic UltraScope, as an ideal companion product.

Nodes

DaVinci invented the term Power Windows, which is often used generically to refer to vignettes and keys added on top of the primary, base layer of color correction. In Resolve, such secondary correction takes place in serial and parallel nodes.  Serial nodes cascade after the primary correction. For instance, you might add windows around three different people in a scene in order to brighten each one. Typically this would involve three nodes (one per person) added after the primary correction to the background scene.

A parallel node is one that is used to extract mask information from any earlier node, in order to apply it later in the chain. By using a parallel node the information isn’t contaminated by other corrections done in the in-between nodes. For example, you might create a parallel node with a key based on an actor’s skin color. This parallel node could be applied downstream as the mask for one node to correct the interior (the person’s skin tones) and a second node for the outside area. Although this might some confusing, it’s the same concept as is used in any node-based compositor.

There are the standard shape and keying tools, but a real hallmark of Resolve is a superb tracker. Simply add a shape around a region to be analyzed and the tracker will automatically create a cloud of tracking points. It is very fast in generating tracking data and was flawless on my test shots.

Although I didn’t test any 3D content, both DaVinci and the Decklink HD Extreme 3D Plus card are ready for stereo 3D projects. This means that the same grade can be applied to left and right eye files and a stereo 3D view can be monitored through the card. The Decklink card will pass a 3Gb/s signal, which will accommodate two HD streams over a single cable.

The test drive

I ran a series of tests using different flavors of Apple ProRes media. In spite of having only the GT120 card, ProResLT at 25fps ran at close to real-time with primary correction. Adding a couple of nodes dropped the playback rate to under 20fps, but still tolerable. On a different testing session, I had two monitors connected and was using standard ProRes media at 29.97fps. Here I noticed more sluggishness as the rate dropped to under 10fps. Obviously, the extra FX4800 card would be a major performance boost, however, rendering was still fast.

A big selling point for Resolve is the ability to take in a single clip of self-contained media for a commercial or show and then use an EDL to frame-accurately parse that clip into individual shots on the Resolve timeline. This isn’t unique, but Resolve does it very well. I wasn’t quite so lucky when I took an FCP project (as individual media clips) into Resolve using an EDL. Quite a number of the shots were linked to the wrong clip. This project originated with a Canon 5D and had some repeating timecode and no actual reel numbers. Although the clips had unique names, Resolve had trouble getting it right. I suspect that using EDLs for file-based projects is still being tweaked at DaVinci.

In spite of being new to the Mac platform, Resolve is already a seasoned product, known for being the DI tool of choice for numerous blockbuster features. This maturity shows in things like a solid integration of RED’s SDK for camera raw REDCODE media files. RED footage enjoys the option of source correction, which can be globally adjusted for an entire project. The raw settings for individual shots can also be directly altered in the timeline, so you can start with an overall setting for all RED clips and then fine-tune individual shots as needed.

Conclusion

Overall, I was happy with the results I got with DaVinci Resolve. It’s a great tool, but mouse operation can be frustrating. Whether or not the toolset is to your liking is strictly a matter of taste. I wish it had a plug-in architecture, as some of the more exotic results you can get by using filter effects in Apple Color or a with a filter like Magic Bullet Looks just aren’t available. There is a built-in blur tool, but it isn’t as smooth nor achieves as extreme of a blur as many other types of Gaussian blur filters.

Blackmagic Design has set out to shake up the high-end world of color grading, by offering the industry’s leading name-brand product at a price range that’s affordable to smaller shops and entrepreneurially-minded colorists. Whether Resolve on the Mac does to color grading what Final Cut Pro did to editing has yet to be seen. Nevertheless, DaVinci Resolve on the Mac puts a product designed by years of input from the world’s leading colorists into the hands of editors and colorists at all levels.

Written for NewBay Media LLC and DV and Videography magazines.

© 2010 Oliver Peters

ARRI ALEXA post, part 3

Post Workflow – Avid Media Composer 5

Thanks to Avid Media Access (AMA), the Media Composer application starting with version 5 can natively handle Apple ProRes media. This means you can bring in files from an ARRI ALEXA and immediately start editing. But ProRes 422 (HQ) or ProRes 4444 might be more taxing than you want to deal with on a long project, so I’m going to outline a suggested workflow for an ALEXA project using Avid Media Composer 5.

Step 1. Same as step 1 in the previous FCP post – clone your media files for protection. Standard archiving solutions include LTO data tape, redundant copies on hard drive or burning media to Blu-ray discs (BR-ROM).

Step 2. Import the camera masters (ProRes HQ or ProRes 4444) into a Media Composer project, using “Link to AMA Files (or Volume)”.

Step 3. Determine the selected takes you plan to edit with and transcode these AMA clips to Avid media. If you intend to use these for just rough cut, creative editing, then select the DNxHD 36 resolution. Move the transcoded files to a separate bin. Close the bin with the AMA clips and do not use these as sources in your edit.

Step 4. Edit as you normally would with the DNxHD 36 clips. Feel free to change clip names as appropriate. I am going to assume that the ALEXA clips used the Log C profile, which means that they are low-contrast in appearance. For the purpose of client/producer viewing, you may need to apply a basic color effect filter to an upper track in order to provide a basic adjustment to make these clips appear more like “normal” video. Edit until the cut is locked and approved.

Step 5. Once the cut is locked, hide the transcoded media by moving the corresponding Avid media files out of the MXF folder on your media hard drive. This will cause the clips in your sequence to be offline.

Step 6. Open the AMA bins (with the ProRes files) and relink the edited sequence to the AMA clips. When relinking is complete, the sequence will be repopulated with AMA media, which will be the native files.

Step 7. Consolidate/transcode your sequence. Select a higher resolution, such as DNxHD 175X. This will create new Avid media clips at full quality for all clips in your timeline.

Click on the image for an enlarged view. (Alexa clip courtesy of ARRI.)

Step 8. Remove any temporary color effect you might have added before. Enter the Avid color correction mode for final grading. As I previously discussed with Color, one approach is to use the auto-contrast controls to change black and white points. Apply this first and increase the master saturation to 150 as a basic starting point. Then proceed to do your creative grading using the hue offset and curve controls. Just as in Color, this automatic control is only calculated on the parked frame, so tweak as needed.

An interesting hybrid approach to posting ALEXA files is to combine elements of these Final Cut and Media Composer workflows. For example, you could use FCP and the Nick Shaw LUT plug-in to create editing proxy files. Export one set of files from FCP with the LUT and timecode burned-in. Bring those in via AMA and follow the steps outlined above for the rough cut.

Then, when it comes to the finishing steps, use AMA to import the original camera files and link your sequence to those clips instead of the ones with the LUT and burn-in. The one caveat to proper relinking is to make sure the source file names match, including extensions. Both the original files and the exported “baked” files need to match with regards to the presence or absence of the .mov file extension. It may take a bit up trial-and-error for you to get the hang of it, but this hybrid solution will work just fine!

©2010 Oliver Peters

ARRI ALEXA post, part 2

Post Workflow – Apple Final Cut Pro

Posting ARRI ALEXA’s ProRes clips is going to follow many of the established workflows. The fact that it’s a new camera doesn’t really make much difference. The camera generates direct-to-edit ProRes media with embedded timecode and reel IDs. These are self-contained files without a folder hierarchy to mess you up, like with P2 or XDCAM. These media files are immediately ready to edit with most modern NLEs, including Apple Final Cut Pro 7, Adobe Premiere Pro CS5 and Avid Media Composer 5 (via AMA). Assuming that these camera clips are intended as the master files, then “best practices” for tapeless workflows need to be followed to archive and protect the media. These files are your “electronic negative”.

You could, of course, immediately edit natively with these clips, but instead, I’m going to lay out an offline-online workflow using ARRI ALEXA clips. Since these master clips might be ProRes 4444 or ProRes (HQ), many editors would rather deal with smaller proxy editing files during the creative, rough cutting phase of a project.

For the sake of this exercise, I’m going to assume that the files are recorded using the ALEXA Log C profile. This yields a low-contrast, “flat” image designed to preserve dynamic range and offer optimum grading latitude. It’s a wonderful alternative to true camera raw recording, but adds a few considerations for post.

Step 1. Clone your media files for protection. Standard archiving solutions include LTO data tape, redundant copies on hard drive or burning media to Blu-ray discs (BR-ROM). Here is also a step-by-step guide from Abel Cine Tech about how to offload the cards for FCP editing.

Step 2. Import the camera masters (ProRes HQ or ProRes 4444) into an FCP project.

Step 3. Drag-and-drop all of your clips to a new sequence. Apply Nick Shaw’s custom ALEXA LUTs to the timeline clips. This is an FxScript FCP plug-in with an optional burn-in window for text and timecode. Use the version of the filter that matches the exposure index to which the camera was set during recording. Now select all the clips in the timeline and drag them to a new, empty bin. This will create subclips with the embedded plug-in (LUT + burn-in).

Step 4. Set up a Batch Export of these subclips. Change the export settings to a different codec, such as ProRes 422 (Proxy), DVCPROHD or anamorphic DV. The exported clips will now become your editing proxy media.

Step 5. Create a new FCP project (optional) and import the folder of exported clips. These will become your working clips for the rough cut. These will have the LUT and burn-in data “baked” into the file.

Step 6. Edit as you normally would until the cut is approved and locked.

Step 7. When the cut is locked, create a new FCP project (optional) and copy the sequence for the locked cut to the new project. Make the sequence clips independent and reconnect the files to the original camera master media. Of course, these clips no longer have the LUT or burn-in applied. Change the sequence settings to match the desired final quality. Adjust any graphics, images and text as needed. Use “send to Color” to move the FCP timeline into Color for final grading.

Step 8. Follow a standard Color grading routine for the sequence. Since there is no corresponding Color LUT for ALEXA files, yet, you will have to establish the correct base look in Color before adding any additional grading. Once the grading is done, render and send back to FCP.

I would recommend, setting a basic conversion grade in Color’s Primary In room that can be applied to all clips. This corrects the Log C profile. Then use the Secondary windows as full-screen “adjustment layers” for all of your normal grading tasks. An alternate approach is to use the Auto Balance control in the Primary In room. I don’t usually think much of such automatics, but it seemed to do a  good job with the ALEXA Log C clips that I tested. Apply it first, which changes the lift/gamma/gain values in the Primary – Advanced tab. Then alter the Primary – Basic saturation value to 1.5. Now you have a good starting point and can still use the curves and hue offset controls (“color wheels”) in the Primary room as you normally would.

Click on the image for an enlarged view. (Alexa clip courtesy of ARRI.)

Remember, that you are not applying a true LUT, since the adjustments you are making – even the automatic ones – are based on the levels for the frame on which you are parked. If a shot moves from bright sun to shade, then setting Auto Balance for the bright sunlit segment will cause the shaded portion to have crushed blacks, as well as the other way around. Be prepared to tweak as needed.

ARRI Digital will soon post a page with a LUTs Generator, which might alter this workflow in the future. Also check out ARRI-specific products at GlueTools.

©2010 Oliver Peters

ARRI ALEXA post, part 1

Dealing with Log C

The introduction of the ARRI ALEXA camera has given the high-end production world a fantastic new digital cinematography tool that combines a variety of recording modes with a robust, production-ready form factor. Most of the details can be found at ARRI Digital, but ALEXA offers HD recording using SDI outputs or on-board Apple ProRes recording. Its large sensor also enables higher resolution recording in the ARRIRAW format, using their proprietary T-link connection to an external recorder. DoPs can have the best of both worlds – camera raw, as well as direct-to-edit recording.

I believe that most crews producing for television projects are going to stick with either on-board ProRes or external video recording options. My guess is that most will skip the ARRIRAW method, unless it’s a project destined for theatrical projection. If a crew opts for external recording, then standard systems like the KiPro, Nanoflash, Cinedeck or an SRW deck will be the logical choices. On-board recordings are stored to two SxS cards, similar to a Sony EX camcorder. ARRI opted to use the Apple ProRes 422 codec family, so the menu can be set to record in any of the five flavors: ProRes 4444, ProRes 422 (HQ), ProRes 422, ProRes 422 (LT) and ProRes 422 (Proxy). In most cases, these would be intended as 1920×1080 master files, which means that typically the operator would set the camera for ProRes 4444 or ProRes 422 (HQ). ARRI has posted a handy camera simulator to help users understand the camera’s menu structure. Another option would be to record master files to an external recorder or in ARRIRAW and use the ProRes on the SxS cards as “dailies” for the editorial team.

A real differentiating factor of the ALEXA is the ability to record with two gamma profiles: Rec709 and Log C. These profiles define how an image will appear on a monitor. HD video is based on the Rec709 standard. Log C is a alternate profile designed to squeeze the full dynamic range of the senor into a video recording without clipping highlights or crushing blacks. It is designed to mimic the profile of 35mm negative after being scanned using the Cineon formatThe logarithmic concept for image profiles stems from film scans, but variations of this idea have been used in other camera implementations, like Sony’s hyper-gamma and Panasonic’s film-rec profiles. None of these are as extreme as Log C, so when users first see an uncorrected Log C image, they are taken aback. Log C images appear as having very low contrast with lower chroma saturation when viewed on a standard video monitor.

Gary Adcock and Art Adams have done nice jobs of dissecting how ALEXA deals with dynamic range and color profiles. Unfortunately, what I’ve found is that the average production and post professional gets very dazed and confused when you start to talk about log space. I’m just a poor country editor, so what follows is a far more simplistic explanation of what Log C is and how to deal with it in post.

Think of the dynamic range that the sensor is exposed to from black to white as a yard stick. When you fit that yard stick diagonally into a box of exactly the right size, then the yard stick remains straight. But the current display range of video is a “smaller box” – so in order to fit the yard stick, you have to cut off one end (clipping highlights) or the other (crushing shadow detail). Or, you can simply push in on the ends of the yardstick, causing it to bow and thereby able to fit into the box. This approach to “push in the ends” and “bow” the yard stick is precisely what a Log C profile is doing. Except that it’s doing it mathematically by remapping luma/chroma/gamma coordinates. The visual manifestation of this profile is a low-contrast, “flat” video image.

In order to correctly “straighten” the curve, you have to apply a LUT (look up table), which alters the values to a new profile based on the target display space. A LUT simply converts the video from one space (the recording) into another (how it is to be displayed). So the conversion LUT for a Rec709 display would be different than the LUT for a digital cinema projector. For better or worse, most video projects are not nearly so precise, which means that the Log C image from an ARRI ALEXA can be turned into a nice, viewable (Rec709) image with very simple tools.

I should interject at this point that I am advocating an approach that is technically incorrect, but works in a practical sense. Most colorists are going to throw their hands up in disgust here, but I’m suggesting a workflow that’s appropriate for many video projects (spots, corporate, web, some TV), but probably shouldn’t be followed when a completely accurate rendition is important. This would most likely be the case when using the ALEXA for a feature film or high-end visual effects. In that instance, you really do want the proper conversion LUTs.

Grading is very subjective and the Log C profile is within an adjustable range. In other words, you can use built-in correction tools and get a very pleasing image, even though that might not be the correct technical approach. This means that if you are working with Log C ProRes files in Avid Media Composer, Apple Final Cut Pro or Adobe Premiere Pro, it’s a simple matter of applying the standard color correction tools to a clip and make it look good in your Rec709 monitoring environment and output to tape.

One good tool is Nick Shaw’s LUT for ALEXA. It’s a FxScript plug-in for FCP that corrects the contrast range and increases chroma saturation. Since his processing is only 8-bit, he disclaims it as to be used for preview purposes only; but quite frankly, I think you could successfully master with it on many projects. It also adds an optional timecode and text burn-in generator, making it an ideal tool to use when delivering “dailies” to a client.

Click on this image to see an enlarged view.

I have done some limited experimentation with ALEXA clips, so this is by no means an exhaustive post on the subject; however, here are my observations thus far. The ARRI ALEXA is capable of a wide dynamic range with very low sensor noise. The Log C profile generally keeps the midrange tones in the right location and pushes the highlights and shadow detail towards the middle. The highlight range tends to be more compressed leaving a lot of range for shadow detail. When you expand this overall range, the midpoint is largely unchanged, which means that you introduce very little noise into the grading process.

When you adjust levels to raise the gamma or add “black stretch” on a typical Rec709 image, you inevitably inject video noise. I don’t see this same sort of anomaly in the few ALEXA clips that I’ve taken through some grading tests. In fact, I’ve done some pretty extreme changes to these clips and they’ve held up quite nicely.

In the next couple of posts, I’ll discuss some editing workflow routines that can be used with ARRI ALEXA projects.

©2010 Oliver Peters