In this Instablog, I will cover technical research into Raystream's technology that I have performed that demonstrates that Raystream's technology performs identically to the freely-available x264 software using the default settings. By doing the these tests, I can provide an objective, quantitative assessment of whether Raystream's technology is any different in practice, instead of much of the paid promotion and uniformed speculation that is currently present around the web.
The investment thesis behind Raystream is that video compression can compress videos to approximately 70% without a significant loss of quality. To begin my research, I downloaded two HD resolution (1080p) trailers from H264Info.com for the movies "I Am Legend" and "The Bourne Ultimatum". The videos are high-resolution, high-bitrate videos ideal for further compression. The original file sizes were 121 MB and 79.6 MB, respectively.
Raystream compresses the videos by 76% and 68%, respectively, although this measure includes the audio, too.
As a comparison, I downloaded the freely-available x264 encoding program and used the default settings to encode the video:
To discern the relative image quality between compression technologies, mathematical measures like Peak Signal-to-Noise can be used, although there are flaws with this formula related to the fact that the human eye's perception of visual quality does not necessarily coincide with PSNR. To remedy this, academic researchers Zhou Wang, Alan C. Bovik, Hamid R. Sheikh and Eero P. Simoncelli created what is known is the Structural SIMilarity (SSIM) index, a method for measuring the similarity between two images. By comparing the pre- and post- compression videos for both x264 and Raystream, we can compute an objective measure of image quality that is not subject to human bias. An SSIM of 1 represents perfect correlation with the uncompressed version, while an SSIM of 0 represents no correlation with the uncompressed version. Ideally, we want SSIM to be as close to 1 as possible.
To increase the sample size, I downloaded three additional HD clips from another website, which show HD video of a shuttle launch, astronauts in space, and a spaceshuttle backflip.
For each video, I imported the original video, the x264 video encoded video using the default encode settings, and the Raystream video into MATLAB for analysis. I used the code available from the researchers' website for the SSIM, and it took only a few lines of code to produce a program capable of analyzing the free product x264 versus the proprietary product "produced" by Raystream: In this way, we can use a mathematical formula to compute an objective measure of Raystream's compression quality versus the default settings of the free encoder, x264.
Discussing their supposed technology, Raystream's website says:
If it is really true that "before Raystream," a compression technology this significant did not exist, then the investment thesis for Raystream is that their video compression technology outperforms other video technologies and therefore their compression will give them an edge over do-it-yourself encoding or other proprietary encoding schemes. The compression technology should be so good that it is worth paying a small company money to license that technology. But as we can see, the Raystream technology appears to make no significant difference over x264 in terms of compression size and image quality:
"Before Raystream, a one hour video converted to 720p using the best compression algorithms resulted in files in excess of 1 GB, far too large to be streamed over commercial Internet connections.
Using Raystream, the same one hour 720p video can be compressed up to 90% of its original file size, which makes it easily streamable over connection speeds ranging from 0.4 to 1.0 Mbs per second."
This table shows the video label, the compression percentage using the default x264 setting, the Raystream compression percentage, the x264 SSIM quality measure, and the Raystream SSIM quality measure. The compression percentages reflect the video-only compression, and it shows that most of the time, there was no need to adjust any bitrates.
For all intents and purposes, the compression numbers that come out are not statistically different from each other. The quality measure of x264 is always slightly better even when x264 compresses just slightly more. For four out of the five videos, the compression percentages and quality measures are nearly identical. This is strong evidence that Raystream has not much or nothing more than x264 running at nearly the default settings for their supposed proprietary technology. One exception was the Bourne Ultimatum trailer, which Raystream compressed about 10% more, but this may be an outlier -- all other videos compressed nearly identically both in terms of compression and image quality. The "I Am Legend" trailer was of similar resolution, bitrate, etc. yet the compression percentage and image quality measure were basically identical for Raystream and x264, so it may be an anomaly in the encoding procedure to produce the one outlier.
In any case, the measurements here appear to show that Raystream's technology does not offer a compelling solution for video compression over the freely-available x264 encoder. The test of this was easy to put together and shows that Raystream's claim that "before Raystream," compressions such as the ones Raystream achieves were not possible, is patently false. Virtually the same compression ratios are achievable using a free product.
In the end, the only thing Raystream appears to offer is an HTTP interface for users to upload their videos and have Raystream do the x264 compression for them. Why would a company pay for this, when I posted the exact default command to tell x264 to do essentially the same compression in-house for free?
In another vein, the company's financials are dire. The latest 10-Q report shows the company's tangible equity is negative, no revenues (aside from $200 leftover from the shell company), and carry 8 million dollars of goodwill as an asset! This trick relies on valuing the purchase of Raystream Gmbh, the German company started to market what appears to be a clone of x264. After what we've seen here, it doesn't seem to me that the purchased company's technology is worth anything, considering it's being put online for free. Again, it appears to me that the only thing Raystream can offer is to do the encoding for its "clients," although it isn't clear to me why a company would not see Raystream merely as a middleman that could be cut out or whose job could be done completely in house.
So that others may run their own tests to reproduce, the MATLAB code used to compute the SSIM values is:
To keep things fair, it compares only the left half of the video for the image quality test. This is because Raystream adds a watermark to their videos in the upper right corner that would lower the image quality measure, so doing this keeps the comparison one of compression technology only.
I hope that by providing the above code, others can perform their own tests using Raystream's technology to confirm that it isn't all it's cracked up to be. The code runs as a script in MATLAB and requires the free "mmread" package for importing the videos into MATLAB, which is available on the Mathworks File Exchange. All videos were compared using between 100 and 400 random frames to give an accurate comparison that doesn't require an extreme amount of memory, since all frames need to be stored decompressed in memory for the SSIM comparison. Further work by others to compare more videos would be a valuable addition to the results presented here.
We have seen how five high bitrate HD videos compressed using Raystream's trial compare closely with the open source x264 encoding technology. The videos were taken from websites where they remain freely available, allowing anyone access to the high-bitrate source whose compression was studied in this Instablog. I also presented the code that will run in MATLAB for anyone to reproduce these measurements and to perform their own. In summary, Raystream appears to offer nothing compelling over a free product and thus my opinion on this stock is a STRONG SELL.
Disclaimer: I currently hold a short position in this stock, and thus I stand to profit from a decline in its share price. I am not a professional investment advisor, but this research represents my honest opinion at the present time. All information provided is for informational purposes only and should not be deemed as investment advice or a recommendation to purchase or sell any specific security. While the information presented herein is believed to be reliable, no representation or warranty is made concerning the accuracy of any data presented. In addition, there can be no guarantee that any projection, forecast or opinion in this document will be realized. All trade names, trade marks, service marks, and logos herein are the property of their respective owners who retain all proprietary rights over their use. I may hold my short position, hold a neutral position, or hold a net long position in the stock after the publication of this article at my own discretion.