The Streaming Toolbox: Cinnafilm PixelStrings, THEOplayer, and Resonance AI
In this installment of The Streaming Toolbox, we'll cover a platform that transforms video formats, another tool that actually looks at what is resonating with viewers using AI and machine learning (ML), and a third that is a player which is delivering the product the first two tools are working with.
Cinnafilm PixelStrings
While most content is produced today in a high-resolution ProRes or DNxHD format, streaming services by their nature need compressed versions, and moving content from one format to another is what Cinnafilm’s new cloud-based platform PixelStrings is all about. This SaaS platform, launched in May 2018, takes the company’s 15 years of experience creating on-prem products (like Tachyon, Dark Energy, and Wormhole) and offers pay-as-you-go cloud options accessible via browser or API for a range of video conversion and optimization.
PixelStrings addresses the growing need for converting different formats, resolutions, data rates and image sizes, at scale. "Anyone can convert, clean, fix, optimize and/or transcode their content. They can think of it as an anything-to-anything conversion platform in the cloud," says Curtis Staples, director of enterprise and OEM partnerships, adding that the company believes it democratizes high-end image conversion. "Frame rate, runtime, resolution, dynamic range … all of those things can be converted in PixelStrings at about 90% less cost than it would typically cost if you had to buy all the gear."
For streaming, PixelStrings can help make better-looking streams with fewer bits, because oftentimes as much as half of content payload is noise. It also handles frame rate conversion, including SD to higher resolutions, HD to UHD, deinterlacing, noise and grain management, broken 3:2 cadence repair, dust busting (removing frame anomalies), retiming content (shortening or lengthening the runtime without changing the pitch of the audio), and Technicolor tone mapping for SDR-to-HDR conversions. Output formats currently include all major professional codecs, plus Interoperable Master Format (IMF). Multiple conversion processes can all be done in one pass.
"There's no end of formats to create content in, and there are hundreds of places to distribute that content. It's very, very rare that the thing that you made either in 1985 or 2005 or yesterday is exactly the format you need to deliver," says Staples.
The platform’s wide-ranging controls are especially helpful for bringing back-catalog content up to current formats, going from broadcast to streaming, and any other variations for multi-platform delivery. The majority of content distributors have never had access to this kind of technology, Staples says, noting, "They just live with broken film cadence and mismatched field dominance and all kinds of weird stuff."
PixelStrings accepts all common professional formats. Media files generally live in Amazon S3 buckets or can be uploaded from local storage. Support for other cloud environments is in the works.
The platform was easy to use for a rank beginner. There are predefined presets as well as the ability to create custom settings. Cinnafilm offers decent help files, as well as instructional videos at vimeo.com/cinnafilm.
Pricing: Free and subscription accounts are available. Costs are calculated on a per-media minute basis, with a minimum base price starting at $1.99 per minute of media and a discount for subscription plans.
THEOplayer
THEO Technologies' Universal Video Player recently won a 2019 Streaming Media Readers’ Choice Award—its third in a row—so it must be doing something right.
THEO's product is a single video playback component solution, providing a similar experience on web, mobile web, native apps, and connected devices. There are two ways to use the player. The first is UI Skinning (demo.theoplayer.com/ui-skinning), says Pieter Liefooghe, SVP, product and operations. "Second, if the service provider wants to create a fully custom experience, then he has to create the UI completely using our APIs and events. Here, the sky is the limit."
New features include DASH low latency with optimized adaptive bitrate (ABR) logic, picture in picture, ad insertion, offline playback, smart TV software development kits (SDKs), and support for the move from client to server-side ad insertion (SSAI). "In SSAI, you splice in ads on the server, but you still need to inform the player about where those ads are to implement ad skip control functionality," says Liefooghe. "Additionally, increasingly advertisers want to see ad playback stats originating from the individual players. As a consequence, these stats (aka beacons) need to be pushed towards an ad tracker from the players." THEO also offers pre-integration with encoding, analytics, advertising, and DRM providers.
Since I’m not a developer, I asked Yazmin Wickham, senior director of digital platform management at Katz Networks/E.W. Scripps Co., for her impressions. "For new services, they have a lot of nice features," she told me. Each service adds "weight" to the player size (literally, in terms of how many bytes of JavaScript code are downloaded by the browser). So, the player can be very "lightweight" in terms of actual bytes of code. She highlighted the following features as notable:
- Autoplay
- Advertising format tester tool
- Related content
- Up next
- Social sharing
- Advertising VMAP (video multiple ad playlist) support
- Frame accurate seeking
- Chapter markers
- Closed captioning and subtitling
- DRM (lots of pre-integrations)
- Chromecast and airplay
- Preloading
- Google dynamic ad insertion
- DRM
Beyond those features, THEOplayer also offers token authentication and analytics "The use of tokens in the playback URLs is fully transparent to the player," says Liefooghe. "The CMS/back end adds tokens to the playback URL. That returns it to the player app who, on its turn, sets it on the THEO player source."
As for analytics, Liefooghe says, "Every single action, whether related to the playback of the main video asset or advertisement is exposed by our player and can be made available to the analytics library." This data feeds an analytics library on the client, which sends those datapoints to an analytics back end. "Our customers can, however, also use the THEO APIs & Events to integrate with others." Analytics can be customized for a variety of platforms, since each platform typically uses its own analytics integration library.
Next up for THEO is the launch of its High Efficiency Streaming Protocol (HESP), with sub-second latency, 100 millisecond zapping time, ABR, bandwidth reduction, and the ability to integrate into existing workflows. Demo accounts are now available.
Pricing: Up to 10,000 impressions per month free, then a charge per 1,000 impressions, starting at €1.15 and topping out at €0.65 once you surpass 5 million impressions. See http://www.theoplayer.com/licensing.
Resonance AI
Resonance AI is an insights platform for media companies to improve both viewer retention and encourage better production choices. Data and insights from dialogue, audio, color, settings, emotions, themes, and social media tracking is compiled and evaluated with AI & ML to see what viewers connect with.
Providing its customers with deep insight to save time and money as they grow their audience is Resonance AI's focus. "How do we actually know what resonates? We are testing millions of combinations, pulling together all that disparate data and modeling it to see what matters," says COO and co-founder Randa Minkarah. In April 2018, the company announced a partnership with ComScore to give local TV stations the ability to use the Resonance AI platform to understand what is creating viewer engagement in news programming. Beyond that, however, Resonance AI is very circumspect about customer data, so while the company wouldn’t share client names, it would say that film studios, advertisers, news stations, and episodic TV creators are among those using this platform.
Resonance AI looks at a wide range of datapoints. Who is the actor or presenter? What emotions are being shared within the content? How quickly is the content moving? What are the themes? Is it a simple or complex scene? What is the excitement level? Does the dialogue complexity resonate with what the audience expects? Resonance AI can even track an actor as he moves through a scene with a mask on. "We’ve developed technology that allows us to track the person, even if the person starts off at the front of the screen and walks all the way to the back," says Minkarah.
Content data is then combined with other metrics like number of times the content/app was downloaded or what level of online engagement it received on social media. "We monitor over 300 million websites on social and search," says Minkarah. "We know that if they're not talking about your show, they're not watching it."
For instance, one TV show featured a few episodes with drastic differences in visual and dialogue complexity. "They lost audience, and we were able to say to them in a very positive way, when looking at writing another season, we would recommend that you tell the same story, however, don't concentrate it so much. That's a practical outcome," she says.
"There are some shows that are very complex because they have many locations, and that increases expenses dramatically. We can say, 'Instead of 11 locations, 9 would actually be equally as productive,'" Minkarah says.
We were not able to demo this product.
Pricing: Based on volume plus additional charges depending on type of analysis. A research group at a small OTT platform would be less than $10,000 per month, while a large content producer with users spanning creative, production, and marketing would be more than $50,000 per month.
Related Articles
In this installment, we look at products for interactive live streams, ad inventory pricing, and CDN balancing
09 Mar 2021
This installment of The Streaming Toolbox is all about remote control, for mobile devices, advertising, and network operations
17 Dec 2020
We look at a B2B intelligence layer for content transport, a company that enables customer relationship management decisioning, and a service that facilitates offline viewing.
29 May 2020
This month, we look at tools for content recognition, file transfer, and low-latency WebRTC streaming
17 Dec 2019