Debunking Some Myths Of The Google/On2 Deal, Questioning VP8’s Quality
Following up on my earlier post today entitled "Google's Acquisition Of On2 Not A Big Deal, Here's Why", here's some more thoughts on the subject. While clearly no one, including me, truly knows what Google plans to do with On2, a lot of potential scenarios being discussed on the web revolve around facts that just aren't accurate. I'm all for having a discussion on what Google may or may not do with the On2 assets, but a lot of folks are using bad info to come up with their logic behind what Google may or may not do. Here are some of the "myths" I keep hearing or reading about:
- Google will no longer have to pay Adobe a license fee for using Flash on YouTube: The reality is that Google does not stream videos from a Flash server using the Flash streaming protocol of RTMP. Since all YouTube videos are delivered via HTTP, progressive download, Adobe is not getting any license fee from Google.
- Adobe has to pay On2 licensing fees for support of the VP6 codec: Adobe paid a one-time fee to license VP6 for Flash years ago. They don't currently pay any royalty for VP6 and there is no reason why Adobe would have to pay Google any kind of license fee to continue to support VP6.
- Google already uses VP6 for YouTube: YouTube has not used VP6 for their videos. They originally started off by using H.263, Spark, and added H.264 support for their HQ and HD videos.
- Google will now be able to "speed up" YouTube videos: Being able to speed up the delivery of video primarily rest in how the content is delivered and the server and protocols used. While compression does play some role in this, the ability to "speed up" the delivery of videos primarily comes from the infrastructure side of the ecosystem, not the encoding. On2 has no server component so even if Googe adopted VP7 or VP8, don't expect the poor buffering times on Google to be fixed.
- Acquring On2 will allow YouTube to save on bandwidth costs: Again, while the rate of compression has an impact on the number of bits being delivered, it is not the biggest factor involved in the costs of delivering video. Yes, if you can compress the file smaller and have better quality, you might be able to save some money, but typically when that happens you increase the bitrate and end up delivering more bits at higher quality. And with YouTube moving to HQ and HD video, they are delivering more bits, not less. They didn't spend $100M to save on bandwidth costs.
- Because of On2's encoding products, it might save Google some money in electricity costs and server hardware: The reality is that On2's transcoding based service runs off of Amazon and I don't see Google cutting checks to Amazon over the long run to keep this service up and running. While On2's Flix encoding software could have some value to YouTube in helping them ingest and transcode videos, the savings they would see, if any, would not warrant the $100M price tag.
- On2 could have a big impact on video quality and delivery of YouTube: Again, it won't have much impact on the delivery aspect. Could it have an impact on the quality? Possibly. But until we see VP8 in action, no one truly knows.
As for the issue surrounding the quality of VP8, we don't know how good the codec is. I have been getting a lot of calls and questions asking me why I think the quality of VP8 is not as good as H.264. I never said the quality was not as good, I said, "Everyone is assuming the quality is better than H.264, but is it?" It may be, but the point is, we don't know. I have spoken to two content owners who were using VP8 under beta and told me there were not impressed. On2 announced VP8 back on September 13th of 2008 and nearly a year later, where is the adoption? Not to mention, the subtitle of the On2 release says "On2 video delivers over 50% bandwidth savings compared to leading H.264 implementations." That may be, but where is the data to prove it?
For nearly the past year, StreamingMedia.com has made multiple requests to On2 to let us get hands on with VP8 so we can compare it for ourselves, just as we did in 2006, when Jan Ozer produced a 100 page report comparing all of the Flash video codecs. To date, we've not been given anything to test with VP8 and it's worth noting that the H.264 sample clips that On2 uses on their website and at tradeshows as comparison to VP8, are not optimized H.264 videos, so the comparison is not fair.
Is VP8 really better than H.264 from a quality perspective? It may or may not be. The point is, let the industry get hands on with it and test it themselves. I'm sure many will say it must be better if Google is willing to acquire On2, but keep in mind we are all speculating that they even want to compete with H.264, which may not be the case. No on seems to be thinking about the potential live streaming aspect for VP8 which for low latency, real-time streams, could work just as good if not better than H.264 for bi-directional video chat applications like Google Talk.
Clearly we all have a lot of unanswered questions and hopefully soon Google will start to lay out a road map on what they plan to do, although considering they don't like to talk about things and the slow rate at which they move, I'm not holding my breath.
For more details on the subject of VP8's quality, see Tim Siglin's post from last year on StreamingMedia.com