Time to Rethink ProRes RAW?

The Apple ProRes RAW codec has been available for several years at this point, yet we have not heard of any professional cinematography camera adding the ability to record ProRes RAW in-camera. I covered ProRes RAW with some detail in these three blog posts (HDR and RAW Demystified, Part 1 and Part 2, and More about ProRes RAW) back in 2018. But the industry has changed over the past few years. Has that changed any thoughts about ProRes RAW?

Understanding RAW

Today’s video cameras evolved their sensor design from a three CCD array for RGB into a single sensor, similar to those used in still photo cameras. Most of these sensors are built using a Bayer pattern of photosites. This pattern is an array of monochrome receptors that are filtered to receive incoming green, red, and blue wavelengths of light. Typically the green photosites cover 50% of this pattern and red and blue each cover 25%. These photosites capture linear light, which is turned into data that is then meshed and converted into RGB pixel information. Lastly, it’s recorded into a video format. Photosites do not correlate in a 1:1 relationship with output pixels. You can have more or fewer total photosite elements in the sensor than the recorded pixel resolution of the file.

The process of converting photosite data into RGB video pixels is done by the camera’s internal electronics. This process also includes scaling, gamma encoding (Rec709, Rec 2020, or log), noise reduction, image sharpening, and the application of that manufacturer’s proprietary color science. The term “color science” implies some type of neutral mathematical color conversion, but that isn’t the case. The color science that each manufacturer uses is in fact their own secret sauce. It can be neutral or skewed in favor of certain colors and saturation levels. ARRI is a prime example of this. They have done a great job in developing a color profile for their Alexa line of cameras that approximates the look of film.

All of this image processing adds cost, weight, and power demands to the design of a camera. If you offload the processing to another stage in the pipeline, then design options are opened up. Recording camera raw image data achieves that. Camera raw is the monochrome sensor data prior to the conversion into an encoded video signal. By recording a camera raw file instead of an encoded RGB video file, you defer the processing to post.

To decode this file, your operating system or application requires some type of framework, plug-in, or decoding/developing software in order to properly interpret that data into a color image. In theory, using a raw file in post provides greater control over ISO/exposure and temperature/tint values in color grading. Depending on the manufacturer, you may also apply a variety of different camera profiles. All of this is possible and still have a camera file that is of a smaller size than its encoded RGB counterpart.

In-camera recording, camera raw, and RED

Camera raw recording preceded the introduction of the RED One camera. These usually consisted of uncompressed movie files or image sequences recorded to an external recorder. RED introduced the ability to record a Wavelet-compressed, 4K camera raw signal at 24fps. This was a movie file recorded onboard the camera itself. RED was granted a number of patents around these processes, which preclude any other camera manufacturer from doing that exact same thing, unless entering into a licensing agreement with RED. So far these patents have been successfully upheld against Sony and Apple among others.

In 2007 – part way through the Final Cut Pro product run – Apple introduced its family of ProRes codecs. ProRes was Apple’s answer to Avid’s DNxHD codec, but with some improvements, like resolution independence. ProRes not only became Apple’s default intermediate codec, but also gained stature as the mastering and delivery codec of choice, regardless of which NLE you were using.

By 2010 Apple was successful in convincing ARRI to use ProRes as its internal recording codec with the introduction of the (then new) line of Alexa cameras. (ARRI camera raw recording was a secondary option using ARRIRAW and a Codex recorder.) Shooting with an Alexa, recording high-quality ProRes files, and posting those directly within FCP or any other compatible NLE created the simplest and smoothest capture-edit-deliver pipeline of any professional post workflow. That remains unchanged even today.

Despite ARRI’s success, only a few other camera manufacturers have adopted ProRes as an internal recording option. To my knowledge these include some cameras from AJA, JVC, Blackmagic Design, and RED (as a secondary file to REDCODE). The lack of widespread adoption is most likely due to Apple’s licensing arrangement, coupled with the fact that ProRes is a proprietary Apple format. It may be a de facto industry standard, but it’s not an official standard sanctioned by an industry standards committee.

The introduction of Apple’s ProRes RAW codecs has led many in the industry to wait with bated breath for cameras to also adopt ProRes RAW as their internal camera raw option. ARRI would obviously be a candidate. However, the RED patents would seem to be an impediment. But what if Apple never had that intention in the first place?

Do we have it all wrong?

When Apple introduced ProRes RAW, it did so in partnership with Atomos. Just like Sony, ARRI, and Panasonic recording their camera raw signals to an external recorder, sending a camera raw signal to an external Atomos monitor/recorder is a viable alternative to in-camera recording. Atomos’ own disagreements with RED have now been settled. Therefore, embedding the ProRes RAW codec into their products opens up that recording format to any camera manufacturer. The camera simply has to be capable of sending a compatible camera raw signal (as data) over SDI or HDMI to the connected Atomos recorder.

The desire to see ProRes RAW in-camera stems from the history of ProRes adoption by ARRI and the impact that had on high-end production and post. However, that came at a time when Apple was pushing harder into various pro film and video markets. As we’ve learned, that course was corrected by Steve Jobs, leading to the launch of Final Cut Pro X. Apple has always been about ease and democratization – targeting the middle third of a bell curve of users, not necessarily the top or bottom thirds. For better or worse, Final Cut Pro X refocused Apple’s pro video direction with that in mind.

In addition, during this past decade or more, Apple has also changed its approach to photography. Aperture was a tool developed with semi-pro and pro DSLR photographers in mind. Traditional DSLRs have lost photography market share to smart phones – especially the iPhone. Online sharing methods – Facebook, Flickr, Instagram, cloud picture libraries – have become the norm over the traditional photo album. And so, Aperture bit the dust in favor of Photos. From a corporate point-of-view, the rethinking of photography cannot be separated from Apple’s rethinking of all things video.

Final Cut Pro X is designed to be forward-thinking, while cutting the chord with many legacy workflows. I believe the same can be applied to ProRes RAW. The small form factor camera, rigged with tons of accessories including external displays, is probably more common these days than the traditional, shoulder-mounted, one-piece camcorder. By partnering with Atomos (and maybe others in the future), Apple has opened the field to a much larger group of cameras than handling the task one camera manufacturer at a time.

ProRes RAW is automatically available to cameras that were previously stuck recording highly-compressed M-JPEG or H.264/265 formats. Video-enabled DSLRs from manufacturers like Nikon and Fujifilm join Canon and Panasonic cinematography cameras. Simply send a camera raw signal over HDMI to an Atomos recorder. And yet, it doesn’t exclude a company like ARRI either. They simply need to enable Atomos to repack their existing camera raw signal into ProRes RAW.

We may never see a camera company adopt onboard ProRes RAW and it doesn’t matter. From Apple’s point-of-view and that of FCPX users, it’s all the same. Use the camera of choice, record to an Atomos, and edit as easily as with regular ProRes. Do you have the depth of options as with REDCODE RAW? No. Is your image quality as perfect in an absolute (albeit non-visible) sense as ARRIRAW? Probably not. But these concerns are for the top third of users. That’s a category that Apple is happy to have, but not crucial to their existence.

The bottom line is that you can’t apply classic Final Cut Studio/ProRes thinking to Final Cut Pro X/ProRes RAW in today’s Apple. It’s simply a different world.

____________________________________________

Addendum

The images I’ve used in this post come from Patrik Pettersson. These clips were filmed with a Nikon Z6 DSLR recording to an Atomos Ninja V. He’s made a a few sample clips available for download and testing. More at this link. This brings up an interesting issue, because most other forms of camera raw are tied to a specific camera profile. But with ProRes RAW, you can have any number of cameras. Once you bring those into Final Cut Pro X, you don’t have the correct camera profile with a color science that matches that model for each any every camera.

In the case of these clips, FCPX doesn’t offer any Nikon profiles. I decided to decode the clip (RAW to log conversion) using a Sony profile. This gave me the best possible results for the Nikon images and effectively gives me a log clip similar to that from a Sony camera. Then for the grade I worked in Color Finale Pro 2, using its ACES workflow. To complete the ACES workflow, I used the matching SLog3 conversion to Rec709.

The result is nice and you do have a number of options. However, the workflow isn’t as straightforward as Apple would like you to believe. I think these are all solvable challenges, but 1) Apple needs to supply the proper camera profiles for each of the compatible cameras; and 2) Apple needs to publish proper workflow guides that are useful to a wide range of users.

©2020 Oliver Peters

ADA Compliance

The Americans with Disabilities Act (ADA) has enriched the lives of many in the disabled community since its introduction in 1990. It affects all of our lives, from wheelchair-friendly ramps on street corners and business entrances to the various accessibility modes in our computers and smart devices. While many editors don’t have to deal directly with the impact of the ADA on media, the law does affect broadcasters and streaming platforms. If you deliver commercials and programs, then your production will be affected in one way or another. Typically the producer is not directly subject to compliance, but the platform is. This means someone has to provide the elements that complete compliance as part of any distribution arrangement, whether it is the producer or the outlet itself.

Two components are involved to meet proper ADA compliance: closed captions and described audio (aka audio descriptions). Captions come in two flavors – open and closed. Open captions or subtitles consists of text “burned” into the image. It is customarily used when a foreign language is spoken in an otherwise English program (or the equivalent in non-English-speaking countries). Closed captions are enclosed in a data stream that can be turned on and off by the viewer, device, or the platform and are intended to make the dialogue accessible to the hearing-impaired. Closed captions are often also turned on in noisy environments, like a TV playing in a gym or a bar.

Audio descriptions are intended to aid the visually-impaired. This is a version of the audio mix with an additional voice-over element. An announcer describes visual information that is not readily obvious from the audio of the program itself. This voice-over fills in the gaps, such as “man climbs to the top of a large hill” or “logos appear on screen.”

Closed captions

Historically post houses and producers have opted to outsource caption creation to companies that specialize in those services. However, modern NLEs enable any editor to handle captions themselves and the increasing enforcement of ADA compliance is now adding to the deliverable requirements for many editors. With this increased demand, using a specialist may become cost prohibitive; therefore, built-in tools are all the more attractive.

There are numerous closed caption standards and various captioning file formats. The most common are .scc (Scenarist), .srt (SubRip), and .vtt (preferred for the web). Captions can be supplied as “embedded” (secondary data within the master file) or as a separate “sidecar” file, which is intended to play in sync with the video file. Not all of these are equal. For example, .scc files (embedded or as sidecar files) support text formatting and positioning, while .srt and .vtt do not. For example, if you have a lower-third name graphic come on screen, you want to move any caption from its usual lower-third, safe-title position to the top of the screen while that name graphic is visible. This way both remain legible. The .scc format supports that, but the other two don’t. The visual appearance of the caption text is a function of the playback hardware or software, so the same captions look different in QuickTime Player versus Switch or VLC. In addition, SubRip (.srt) captions all appear at the bottom, even if you repositioned them to the top, while .vtt captions appear at the top of the screen.

You may prefer to first create a transcription of the dialogue using an outside service, rather than simply typing in the captions from scratch. There are several online resources that automate speech-to-text, including SpeedScriber, Simon Says, Transcriptive, and others. Since AI-based transcription is only as good as the intelligibility of the audio and dialects of the speakers, they all require further text editing/correction through on online tool before they are ready to use.

One service that I’ve used with good results is REV.com, which uses human transcribers for greater accuracy, as well as offering on online text editing tool. The transcription can be downloaded in various formats, including simple text (.txt). Once you have a valid transcription, that file can be converted through a variety of software applications into .srt, .scc, or .vtt files. These in turn can be imported into your preferred NLE for timing, formatting, and positioning adjustments.

Getting the right look

There are guidelines that captioning specialists follow, but some are merely customary and do not affect compliance. For example, upper and lower case text is currently the norm, but you’ll still be OK if your text is all caps. There are also accepted norms when English (or other) subtitles appear on screen, such as for someone speaking in a foreign language. In those cases, no additional closed caption text is used, since the subtitle already provides that information. However, a caption may appear at the top of the screen identifying that a foreign language is being spoken. Likewise, during sections with only music or ambient sounds, a caption may briefly identifying it as such.

When creating captions, you have to understand that readability is key, so the text will not always run perfectly in sync with the dialogue. For instance, when two actors engage in rapid fire dialogue, each caption may stay on longer than the spoken line. You can adjust the timing against that scene so that they eventually catch up once the pace slows down. It’s good to watch a few captioned programs before starting from scratch – just to get a sense of what works and what doesn’t.

If you are creating captions for a program to run on a specific broadcast network or streaming services, then it’s a good idea to find out of they provide a style guide for captions.

Using your NLE to create closed captions

Avid Media Composer, Adobe Premiere Pro, DaVinci Resolve, and Apple Final Cut Pro X all support closed captions. I find FCPX to be the best of this group, because of its extensive editing control over captions and ease of use. This includes text formatting, but also display methods, like pop-on, paint-on, and roll-up effects. Import .scc files for maximum control or extract captions from an existing master, if your media already has embedded caption data. The other three NLEs place the captions onto a single data track (like a video track) within which captions can be edited. Final Cut Pro X places them as a series of connected clips, like any other video clip or graphic. If you perform additional editing, the FCPX magnetic timeline takes care of keeping the captions in sync with the associated dialogue.

Final Cut’s big plus for me is that validation errors are flagged in red. Validation errors occur when caption clips overlap, may be too short for the display method (like a paint-on), are too close to the start of the file, or other errors. It’s easy to find and fix these before exporting the master file.

Deliverables

NLEs support the export of a master file with embedded captions, or “burned” into the video as a subtitle, or the captions exported as a separate sidecar file. Specific format support for embedded captions varies among applications. For example, Premiere Pro – as well as Adobe Media Encoder – will only embed captioning data when you export your sequence or encode a file as a QuickTime-wrapped master file. (I’m running macOS, so there may be other options with Windows.)

On the other hand, Apple Compressor and Final Cut Pro X can encode or export files with embedded captions for formats such as MPEG2 TS, MPEG 2 PS, or MP4. It would be nice if all these NLEs supported the same range of formats, but they don’t. If your goal is a sidecar caption file instead of embedded data, then it’s a far simpler and more reliable process.

Audio descriptions

Compared to closed captions, providing audio description files is relatively easy. These can either be separate audio files – used as sidecar files for secondary audio – or additional tracks on the delivery master. Sometimes it’s a completely separate video file with only this version of the mix. Advanced platforms like Netflix may also require an IMF (Interoperable Master Format) package, which would include an audio description track as part of that package. When audio sidecar files are requested for the web or certain playback platforms, like hotel TV systems, the common deliverable formats are .mp3 or .m4a. The key is that the audio track should be able to run in sync with the rest of the program.

Producing an audio description file doesn’t require any new skills. A voice-over announcer is describing any action that occurs on screen, but which wouldn’t otherwise make sense if you were only listening to audio without that. Think of it like a radio play or podcast version of your TV program. This can be as simple as fitting additional VO into the gaps between actor/host/speaker dialogue. If you have access to the original files (such as a Pro Tools session) or dialogue/music/effects stems, then you have some latitude to adjust audio elements in order to fit in the additional voice-over lines. For example, sometimes the off-camera dialogue may be moved or edited in order to make more space for the VO descriptions. However, on-camera/sync dialogue is left untouched. In that case, some of this audio may be muted or ducked to make space for even longer descriptions.

Some of the same captioning service providers also provide audio description services, using their pool of announcers. Yet, there’s nothing about the process that any producer or editor couldn’t handle themselves. For example, scripting the extra lines, hiring and directing talent, and producing the final mix only require a bit more time added to the schedule, yet permits the most creative control.

ADA compliance has been around since 1990, but hasn’t been widely enforced outside of broadcast. That’s changing and there are no more excuses with the new NLE tools. It’s become easier than ever for any editor or producer to make sure they can provide the proper elements to touch every potential viewer.

For additional information, consult the FCC guidelines on closed captions.

The article was originally written for Pro Video Coalition.

©2020 Oliver Peters

Video Technology 2020 – Shared Storage

Shared storage used to be the domain of “heavy iron” facilities with Avid, Facilis, and earlier Apple Xserve systems providing the horsepower. Thanks to advances in networking and Ethernet technology, shared storage is accessible to any user. Whether built-in or via adapters, modern computers can tap into 1Gbps, 10Gbps, and even higher, networking speeds. Most computers can natively access Gigabit Ethernet networks (1Gbps) – adequate for SD and HD workflows. Computers designed for the pro video market increasingly sport built-in 10GbE ports, enabling comfortable collaboration with 4K media and up. Some of today’s most popular shared storage vendors include QNAP, Synology, and LumaForge.

This technology will become more prolific in 2020, with systems easier to connect and administer, making shared storage as plug-and-play as any local drives. Network Attached Storage (NAS) systems can service a single workstation or multiple users. In fact, companies like QNAP even offer consumer versions of these products designed to operate as home media servers. Even LumaForge sells a version of its popular Jellyfish through the online Apple Store. A simple, on-line connection guide will get you up and running, no IT department required. This is ideal for the individual editor or small post shop.

Expect 2020 to see higher connection speeds, such as 40GbE, and NAS proliferation even more widespread. It’s not just a matter of growth. These vendors are also interested in extending the functionality of their products beyond being a simple bucket for media. NAS systems will become full-featured media hubs. For example, if you an Avid user, you are familiar with their Media Central concept. In essence, this means the shared storage solution is a platform for various other applications, including the editing software. There are additional media applications that include management apps for user permission control, media queries, and more. Like Avid, the other vendors are exploring similar extensibility through third-party apps, such as Axle Video, Kyno, Hedge, Frame.io, and others. As such, a shared network becomes the whole that is greater than the sum of its parts.

Along with increased functionality, expect changes in the hardware, too. Modern NAS hardware is largely based on RAID arrays with spinning mechanical drives. As solid state (SSD) storage devices become more affordable, many NAS vendors will offer some of their products featuring RAID arrays configured with SSDs or even NVMe systems. Or a mixture of the two, with the SSD-based units used for short-term projects or cache files. Eventually the cost will come down enough so that large storage volumes can be cost-effectively populated with only SSDs. Don’t expect to be purchasing 100TB of SSD storage at a reasonable price in 2020; however, that is the direction in which we are headed. At least in this coming year, mechanical drives will still rule. Nevertheless, start looking at some percentage of your storage inventory to soon be based on SSDs.

Click here for more on shared storage solutions.

Originally written for Creative Planet Network.

©2020 Oliver Peters

Video Technology 2020 – The Cloud

The “cloud” is merely a collection of physical data centers in multiple locations around the world – not much different than a small storage center you might have. Of course, they employ more advanced systems for power, redundancy, and security than you do. When you work with one of the companies marketing cloud-based editing or a review-and-approval service, like Frame.io or Wipster, they provide the user-facing interface, but are actually renting storage space from one of the big three cloud providers – Google, Amazon, or Microsoft.

There are three reasons that I’m skeptical about ubiquitous, cloud-based editing (with media at native resolutions) in the short term: upload speeds, cost, and security.

Speed

5G (fifth generation wireless) is the technology predicted to offer adequate speeds and low latency for native 4K (and higher) media. While 5G will be a great advancement for many things, it’s a short distance signal requiring more transmission spots than current wireless technology. Full coverage in most metro areas, let alone widespread geographical coverage worldwide, will take many years to fully deploy. Other than potential camera-to-cloud uploads of proxy media in the field, 5G won’t soon be the killer solution. Current technology still dictates that if you want the fastest possible upload speeds for large amounts of data, then you have to tap as close as possible to the internet’s backbone.

Cost

Cloud storage is cheap, but extensive upload and download times aren’t. Unfortunately modern video resolutions also result in huge amounts of data generated on every shoot. Uploading native 4K media for a week-long production is considerably more expensive than FedEx and overnight charges to ship drives. What about long term storage? Let’s say that all of your native media is in the cloud and you pay according to a monthly or annual subscription plan. But what if you want to stop? That media will have to be downloaded and stored locally, which will incur data rate charges, as well as your time to download everything.

Security

Think these sites are unequivocally secure? Look at any data hack at a major company. Security is such a concern in our business that most major movie studios won’t let their editors connect the computers to the internet. Many make these editors check their cell phones at the door. No matter how secure, it’s going to be a hard sell, except for limited slices of the production, such as cloud-based VFX rendering.

I do believe 2020 will be a year in which many will take advantage of some modes of long distance, cloud-based edit services using low-res proxy media. Increasingly some services will be used to move dailies and deliverables around the globe via the cloud. But that’s a big difference from cloud-based editing becoming the norm. One edit scenario many will experiment with is to store the edit project files in the cloud, but with the media mirrored locally at each edit site. This way only the lightweight files used for edit collaboration need be moved over the internet. Think of this as Google Docs for editing. Adobe already offers a version of this, but I suspect you’ll see others, including solutions for Final Cut Pro X. So while true cloud-based editing is not a near-term solution, bits and pieces will become increasingly commonplace.

Originally written for Creative Planet Network.

©2020 Oliver Peters

Video Technology 2020 – Apple and the PC Landscape

Apple enjoys a small fraction of the total computer market, yet has an oversized influence on video production and post. Look anywhere in our business and you’ll see a high percentage of Apple Mac computers and laptops in use by producers, DITs, editors, mixers, and colorists. This has influenced the development and deployment of certain technologies, such as optimization for Metal, Thunderbolt i/o, ProRes codecs, and more. This may irritate Windows users, but it’s something companies like Avid, Adobe, and others cannot ignore. Apple deprecates OpenGL, OpenCL, and CUDA in favor of Metal, and so, developers of software for Apple computers will follow suit so that their Mac-based customers enjoy a good experience.

Going into 2020, Apple is offering a better line-up of professional Mac products than it has in years. MacBook Pro laptops, iMacs and iMac Pros, and the new Mac Pro are clearly targeted at the professional customer. Add to this the Pro Display XDR and authorized third-party products available through Apple, like LumaForge Jellyfish storageBlackmagic and Sonnet eGPUs. Clearly Apple intends to offer an end-to-end hardware and software ecosystem designed to appeal to the pro video customer.

Apple’s prices can be a turn-off for some. Similar investments in a PC – especially custom configurations – may yield better performance in certain applications. Nevertheless, most former and present owners of Mac Pro “cheese grater” towers feel like they got their money’s worth and will at least have interest in the new Mac Pro. Same for MacBook Pro owners. So while these new machines may not move the needle for the larger consumer computer market, it will definitely keep current Mac users in the fold and prevent migration to Windows or Linux PCs. It also reinforces Apple’s interest in the professional market – not just video, but also animation, design, audio, science, and engineering.

The unknown will be the impact of Apple’s new Afterburner card for the Mac Pro. While accelerator cards have been offered by various manufacturers in the past, recent computing developments have focused on processor core counts and GPU technology. The Apple Afterburner is the first introduction for Apple of a new FPGA-based (programmable ASIC) hardware accelerator card. Designed for transcoding, it promises to increase stream counts with 4K and 8K raw and standard codecs in the Mac Pro. Once it’s out in the wild, we will have a better idea of who supports it (beyond Apple’s own software) and its real-world performance.

As Apple goes, so goes the rest of the industry. How will the PC world counter this? Will we see similar cards from HP or Dell? Or will NVIDIA respond with similar results using their GPUs? That’s unknown right now, but my guess is that it will take at least this next year for the rest of the world to respond with competing solutions.

Originally written for Creative Planet Network.

©2020 Oliver Peters