Hey Guys, I'd love to hear feedback on test 3 for smooth streaming on Channel 9.  With IIS Smooth Streaming Beta 2 and a bunch of encoding tweaks, it *should* be vastly improved from the first test.  This is about the video, not about the player features.  

I have a bunch of different video types.  The one video with text right away is the MEF & Silverlight 4 Beta video.  If you're worried about how text will look, check out that video.  If there is a visual issue, please put in a screen shot so I can see the visual issue.  clint.rutkas@microsoft.com

http://smooth.ch9.ms/

For the player, I hear you about wanting the ability to switch to buffering so that feedback is not lost.

With what I assume is the lowest quality the text is pretty much unreadable, but it's fine for the non-text videos. One step up from that is perfectly fine for all but the smaller text on his desktop.

With my crappy connection here at work, I'm very impressed that it starts playing as fast as it does and can manage decent quality.

C

spivonious said:

Much improved! I had zero buffering screens, fairly good quality once it got going (about 1-2 seconds in) and could jump around the videos pretty much instantaneously.

 

With what I assume is the lowest quality the text is pretty much unreadable, but it's fine for the non-text videos. One step up from that is perfectly fine for all but the smaller text on his desktop.

 

With my crappy connection here at work, I'm very impressed that it starts playing as fast as it does and can manage decent quality.

w00t!

Clint said:

spivonious said:

*snip*

w00t!

No bandwith issues here.

Current speed is 3.19mb/s.

I have a question about the technology:

# How far ahead does it buffer? I mean lets say I start watching a 50 minute charles video and haveway stop watching (for some reason) and close my browser. Will it have allready buffer that other 25 minutes? Or will it only buffer one minute ahead (and will i see it buffering like i see the current video's buffer with more brith white seek bar?

# Sometimes I watch the video and afterwards think. This is a really good video this is one I'll need to watch again I'll download it so I have it available for a time I don't have an internet connection. But then downloading it requires me to 'sacrifice' another 500Mb of file transfer althoug I'm sure I have it somewhere downloaded in a temp form of silerlight storage. It there any way to 'save' the stream after watching? Thus not downloading the sae 500mb again (since i download WMV high and I guess that is the file played in the silerlight player)

PS sorry if I rambeld on...

CKurt said:

Very Very good! The first load of the player itself take a while because the app is loading but the video loads very well and seeking works like perfection! Very Very good experience! Also with chapters comming up it will be heaven!

 

I have a question about the technology:

# How far ahead does it buffer? I mean lets say I start watching a 50 minute charles video and haveway stop watching (for some reason) and close my browser. Will it have allready buffer that other 25 minutes? Or will it only buffer one minute ahead (and will i see it buffering like i see the current video's buffer with more brith white seek bar?

 

# Sometimes I watch the video and afterwards think. This is a really good video this is one I'll need to watch again I'll download it so I have it available for a time I don't have an internet connection. But then downloading it requires me to 'sacrifice' another 500Mb of file transfer althoug I'm sure I have it somewhere downloaded in a temp form of silerlight storage. It there any way to 'save' the stream after watching? Thus not downloading the sae 500mb again (since i download WMV high and I guess that is the file played in the silerlight player)

 

PS sorry if I rambeld on...

1>  It will read to whatever the time buffer is set to, the default is 30 seconds.  Also overriding this will cause side effects for stream switching.  And even if the stream is paused, the given stream currently giving data is based off computer, player size and bandwidth qulaity.  So even if it is paused and buffering, you won't get a 2mb stream if the player thinks you can't play a 2mb stream or don't need a 2mb stream for the player size.  A big thing with adaptive streaming is why give you a massive stream if the player size can't even display all that data.

2>  If you want to rewatch the video and worried about bandwidth, I'd lean toward downloading it.  I can put in a feature bug into the player to explore if I can do that on either buffered or adaptive streaming.  I'd also feel more confident in being able to save a buffered stream than an adaptive stream but I'm not sure if this is even possible.

Mike Taulty's video (MEF & Silverlight 4 Beta) is pretty pixelated at the beginning, but it gets better in 10 seconds or so, and becomes crystal clear after just 15-25 seconds.

Quite impressed.

Blue Ink said:

It's really good.

 

Mike Taulty's video (MEF & Silverlight 4 Beta) is pretty pixelated at the beginning, but it gets better in 10 seconds or so, and becomes crystal clear after just 15-25 seconds.

 

Quite impressed.

Same experience as me, blocky and pixellated to begin with, but much better quality later. I get about 120-130ms ping to smooth.ch9.ms btw.

But my issues with C9 were rarely with the streaming (and when I did have streaming issues it was because I had a dodgy connection), they were mainly with the player. Speaking of which, are video streams resumable? If I'm watching over wifi and lose my connection, does the player have to redownload it all when I get reconnected or will it keep its buffer and play what's been downloaded?

The http://www.iis.net/media/experiencesmoothstreaming demo only takes 5 to 9 seconds to get good and clear.

I have a 15mbit cable connection.

Blue Ink said:

It's really good.

 

Mike Taulty's video (MEF & Silverlight 4 Beta) is pretty pixelated at the beginning, but it gets better in 10 seconds or so, and becomes crystal clear after just 15-25 seconds.

 

Quite impressed.

Can you email me a screenshot?  clint.rutkas@microsoft.com

W3bbo said:

Blue Ink said:

*snip*

Same experience as me, blocky and pixellated to begin with, but much better quality later. I get about 120-130ms ping to smooth.ch9.ms btw.

 

But my issues with C9 were rarely with the streaming (and when I did have streaming issues it was because I had a dodgy connection), they were mainly with the player. Speaking of which, are video streams resumable? If I'm watching over wifi and lose my connection, does the player have to redownload it all when I get reconnected or will it keep its buffer and play what's been downloaded?

With adaptive, it only grabs what it needs.  It starts out thinking you're a low bandwidth user then ramps up as needed.  If you can't support a stream, it puts you on a lower one.  The end goal is instant streaming with zero hiccups.

On network failure, with the player as shown, it doesn't not appear to fail gracefully. 

If you refresh and jump to where you left off on both buffered and adaptive, you shouldn't download the area you jumped.

dentaku said:

It took almost 30 sec for the MEF & Silverlight 4 Beta video to get totally clear then I tried it again later and it only took 15 sec before it jumped into good quality. I'm not sure if the other videos had the same problem because they don't have small text in them.

The http://www.iis.net/media/experiencesmoothstreaming demo only takes 5 to 9 seconds to get good and clear.

I have a 15mbit cable connection.

it all depends on the data streams getting sent.  If the player thinks it isn't getting enough to sustain a certian bitrate stream, it will drop down.

For the two different experiences, the network connection could have been saturated at one point and not the another point in time.

Clint said:

dentaku said:

*snip*

it all depends on the data streams getting sent.  If the player thinks it isn't getting enough to sustain a certian bitrate stream, it will drop down.

 

For the two different experiences, the network connection could have been saturated at one point and not the another point in time.

Those videos work much better than what is currently implemented I want it now. It was smooth fast and I could skip ahead with little delay. Currently the silverlight player just freezes

However, I still am not in favour of this. The current, non-smooth streaming videos start just as fast, and have high quality right from the start. There is no argument why we'd even need smooth streaming with my connection.

Sven Groot said:

It's definitely better than the previous two attempts. It starts off sort of blocky but after only about 3-5 seconds it becomes much better. Jumping around is quick, but can occasionally cause it to fall back to a lower quality stream again after which it's very reluctant to go back up.

 

However, I still am not in favour of this. The current, non-smooth streaming videos start just as fast, and have high quality right from the start. There is no argument why we'd even need smooth streaming with my connection.

Some of the current videos are too blurry for me to see code or text on a screen.

Charles said:

Looks great. Very quick seek and the quality is good. Keep on pushing!

C

I can only confirm that it seems to be working very smoothly. One of the best experiences ever for viewing video content.

giovanni said:

Charles said:

*snip*

I can only confirm that it seems to be working very smoothly. One of the best experiences ever for viewing video content.

Likewise

I'm connected via cable modem -> WiFi -> ThinkPad (Intel wireless chip)

I will try this again at various times during the week to see if I get a consistent experience.

Thanks.

Ian2 said:

Living in the countryside my connectivity to the Internet has always been very poor.  This was a revelation - I watched the whole of Mike Taulty's MEF intro without any buffereing!  (and although not Hi Res, all perfectly watchable!).  Genuinly useful.

 

I will try this again at various times during the week to see if I get a consistent experience.

 

Thanks.

Seemed quite blocky here -- on 1.5 (ish) megabit -- OK for talking heads, but not so good for screencasts where I'd want to be able to read the text.  Didn't seem to get less blocky over time.  Will try again this evening.

Herbie

Dr Herbie said:

Ian2 said:

*snip*

Seemed quite blocky here -- on 1.5 (ish) megabit -- OK for talking heads, but not so good for screencasts where I'd want to be able to read the text.  Didn't seem to get less blocky over time.  Will try again this evening.

 

Herbie

 

Looks great here...

on a wired 100/100mbit connection in sweden. jumping around is snappy though. i know its not about the player but a bandwidth meter would help us helping you

Sven Groot said:

It's definitely better than the previous two attempts. It starts off sort of blocky but after only about 3-5 seconds it becomes much better. Jumping around is quick, but can occasionally cause it to fall back to a lower quality stream again after which it's very reluctant to go back up.

 

However, I still am not in favour of this. The current, non-smooth streaming videos start just as fast, and have high quality right from the start. There is no argument why we'd even need smooth streaming with my connection.

As Clint said, smooth streaming is not for everybody.

For the player, I hear you about wanting the ability to switch to buffering so that feedback is not lost.

Thank Clint for answering all the questions about the technology! I'm happy to hear my bandwidth does not go lost if I pauze the video (unlike any other video service out there)

CKurt said:

Sven Groot said:

*snip*

 

Thank Clint for answering all the questions about the technology! I'm happy to hear my bandwidth does not go lost if I pauze the video (unlike any other video service out there)

Looks great here...  starts instantly and quickly (probably within first 10-15 seconds) ramps up to the higher-bandwidth stream.

Nice work; can't wait until this is implemented everywhere on C9!

Using Smooth Streaming here is a good move!!

intelman said:

Clint said:

*snip*

Those videos work much better than what is currently implemented I want it now. It was smooth fast and I could skip ahead with little delay. Currently the silverlight player just freezes

This is why I'm testing out the tech and we're building in the toys needed

Bass said:

The quality seems a little lower then the regular videos on the site. More artifacty.

All depends what stream you are viewing at.  The idea is the player will grab what is the highest possible for that player size.  Currently we send a WMV to you that is far higher quality.  So if you watch the video non-full screen, you are getting data you don't want.  However the plus side to this is you can then automatically transition into full screen and have what will appear to be zero quality loss.  With smooth streaming, it will take a few for you to step up in quality.

Different tech for different purposes.  Each have advantages and disadvantages.

I'd love to see a screen shot, know your geo, and what your internet connection type is, if at all possible.  clint.rutkas@microsoft.com

aL_ said:

looks good, but i think im not reaching full bandwidth on the bach vid for some reason.. blockyness clears up in ~20 seconds for me

on a wired 100/100mbit connection in sweden. jumping around is snappy though. i know its not about the player but a bandwidth meter would help us helping you

Doing a bandwith meter is "interesting" ... I have two different solutions, one I'm not thrilled with, other doesn't work properly for a few reasons.  Smooth Streaming Beta 2 may have fixed some of the issues I was having with the second one.  The big issue was the difference between normal and full screen.

Sven Groot said:

It's definitely better than the previous two attempts. It starts off sort of blocky but after only about 3-5 seconds it becomes much better. Jumping around is quick, but can occasionally cause it to fall back to a lower quality stream again after which it's very reluctant to go back up.

 

However, I still am not in favour of this. The current, non-smooth streaming videos start just as fast, and have high quality right from the start. There is no argument why we'd even need smooth streaming with my connection.

like CKurt mentioned, the feedback about progressive (buffered) playback is not lost.  This test for Channel 9 to see do people like it, does it work for them, do they like the trade offs.

aldenml said:

Hi Clint, I think your are not coding to the highest quality level possible. For example, in RobbieBCES2010, I can easily get the 1605000 bitrate but the edges are extremely aliased compared to the original video. Is there a reason for that? Why not to code the 2000000+ levels?

 

Using Smooth Streaming here is a good move!!

can you send me a screen shot where I can see the time bar?

Clint said:

aldenml said:

*snip*

can you send me a screen shot where I can see the time bar?

Ok, retested.

On a 1.45 MBit link the quality did improve after about 20-30 seconds, but when viewed in full-screen the text was poor.  So it's perfectly fine for talking heads but for a screencast I'd want the current, buffered system instead; I'd rather be patient and get higher quality video.

Looks like we might need both ways.

Herbie

Dr Herbie said:

Clint said:

*snip*

Ok, retested.

On a 1.45 MBit link the quality did improve after about 20-30 seconds, but when viewed in full-screen the text was poor.  So it's perfectly fine for talking heads but for a screencast I'd want the current, buffered system instead; I'd rather be patient and get higher quality video.

Looks like we might need both ways.

 

Herbie

I hear you on that, I have some POCs

Roeland said:

First 10 seconds are not so good, then it starts getting better. Good stuf, keep going!

part of adaptive smooth streaming.  You have trade offs.  It ramps up the quality but this ramp up allows for instant playback and zero buffering.