Review: Bitcodin, a Cloud Video Encoding Service From Bitmovin
In terms of encoding accuracy, Bitmovin averaged about 6 percent over the target data rate when I encoded using the five outputs in the FullHD profile. This compared to 9 percent for Amazon when I encoded to the same parameters. For perspective, both results are actually pretty good; you start to get worried when encoders miss by 15 percent or more.
My final tests related to HLS compatibility and compliance. To test this, I uploaded HLS output to my website and played the streams on several iOS devices, including an iPhone 6 and iPad 1 and 3, with no issues. I then tried playing the streams in Google Chrome on my Samsung Nexus 10 Android tablet, and they didn’t play. I then tried with the HLS streams generated by Amazon with the same result on all test platforms, so it’s hard to point a finger at Bitcodin for the Android problems.
Then I tested both streams with Apple’s Mediastream Validator tool. While Amazon sailed through with no issues, the Bitcodin streams generated the warnings shown in Figure 4 and opened a can of worms regarding the optimal segment size for HLS production. Among the issues identified by the tool were that certain segments exceeded the target bandwidth, and that several media segments exceeded the target duration.
Figure 4. Issues raised by Apple’s Mediastream Validator tool
Note that you can’t choose a segment size in the Bitcodin user interface or public API, and that by default, Bitcodin applies a variable segment size to its HLS/DASH encoded files. That’s the can of worms. To be clear, the segment duration is applied identically to all files in the adaptive group, so stream switching should still be smooth across the group.
As justification for its variable segment size approach, Bitcodin shared research available at that led its engineers to conclude, “Based on the results of these evaluations as well as our experiences from customer deployments, Bitmovin would recommend to use segment sizes around 2 to 4 seconds.”
Of course, this flies in the face of Apple’s nearly universally followed recommendation in Tech Note TN2224, which states, “We strongly recommend a 10 second target duration. If you use a smaller target duration, you increase the likelihood of a stall.” (Emphasis in original.) As a side note, Apple did temporarily switch the recommended segment size to 9 seconds, which I reported last April, but has apparently switched back to 10 seconds. Interestingly, Bitcodin didn’t follow those recommendations, deploying disparate segment sizes within a single 3-minute HLS stream. It’s heresy, I tell you! Heresy!
When I asked about HLS configuration, a Bitmovin representative pointed me toward the aforementioned post and said, “We have already tested our HLS streams with various devices and with a live deployment with one of our customers. They work fine.”
That’s certainly true, as you’ll read about in the sidebar, and the customer referenced (Flimmit) has gotten Apple approval for their app, so you can rest easy on that score. Still, the representative says, users should be able to choose a custom segment size sometime in the future, which sounds like a capital idea.
Overall, while there were a couple of rough edges, Bitmovin is moving in the right direction, and the performance and quality were impressive. Bitmovin needs to (and will) expose more options in the UI and simple interface, such as HLS segment length, enable HLS-only output (without DASH), and add support for MP4 file output. Meanwhile, the user interface is exceptionally simple to use, which will appeal to encoding novices, while the public API should be equally simple to implement for high-volume users; Bitcodin should be very competitive upon its final release.
Flimmit is an Austrian distributor of videos that uses Bitmovin’s Bitdash client and Bitcodin cloud encoding. Flimmit CTO Walter Huber was kind enough to answer some questions about the service and cloud encoding in general. By way of background, Flimmit originally developed its own ffmpeg-based encoder as an Amazon Webservice. It stores master files and encoded assets on S3 and delivers the streams via Cloudfront via RTMPe and HTTPS.
Here’s the Q&A.
Why did you switch to Bitcodin?
We wanted to switch to MPEG-DASH and HLS to move completely to HTTP traffic. We searched for low-cost, ready-to-use solutions in AWS. During my first research on DASH I found some articles from Christian Timmerer (Bitmovin CIO), so I got in touch with him.
How is Bitcodin’s encoding speed?
It is really fast, as we had no bottleneck for our relaunch, where we had to encode more than 4,000 assets in a few weeks.
What about quality?
We are very satisfied with the output quality. Bitcodin has extensive encoding experience and resolved any problems we experienced within 1 business day.
Are you encoding via the API or UI?
We only use the API.
About how long did it take to create the interface between your systems and Bitcodin?
We created the interface very quickly, as the basic API is really easy to use. Based on this experience, we requested features like additional encoding settings and some workflow-related features that Bitmovin promptly delivered. So this was a continuous process during the last months.
Can you share how many streams you create for your typical SD and HD adaptive groups?
This is our internal magic, which we won’t share.
If you were advising a colleague considering a cloud encoder, what should be their most important criteria for choosing a cloud encoder?
First of all cost-effectiveness, then an easy-to-use API combined with extended encoding settings, a quick-start API for easy use, a simple user interface, broad format support, good encoding speed, perfect output quality.
This article appears in the Summer 2015 issue of Streaming Media European Edition as "Review: Bitcodin.”
Patchwork video workflows happen when companies add solutions from different providers, and they can seriously hamper quality and development.
The benefits of per-title optimization aren't just for the major players, anymore. Streaming Media reviews the first solution for smaller content owners and finds the results promising.
Customers can demo the DASH players online and even use a free trial version on any noncommercial website.
Two years from its formation, Austrian startup Bitmovin will grow its product line and expand in Europe and the U.S. with seed funding.