Sony Vegas 13 Sony AVC vs MainConcept - My tests, your experiences.
This is my first post over here, and I hope I have researched most things well enough thanks to all the informative posts here on creative cow.
I did some render tests last night, to find the best way to render my videos. I'm currently using an i5 4690k along with a gtx 960 (so unfortunately I can't make use of this GPU for rendering.) I'm mainly working with 720p 60fps footage for video games, and 1080p 30 fps for videos filmed with my 600D/t3i.
The video I tested on, did have a few effects on including: NeatVideo, Noise to dither the video, Sony Convolution Kernel to sharpen, Magic Bullet Looks (cinestyle s-curve & the shoulder tool) and Sony Levels Computer RGB to Studio RGB.
I recorded 1 minute of video, and rendered them with the following settings:
-MainConcept default settings single pass
-MainConcept default settings 2 pass
-Then I dropped the max bitrate to 10mbps and to min 8mbps again with 2 pass
-And finally MainConcept CBR at 8mpbs (per YouTube's guideline.)
Each test took around 19 minutes despite the changes in bitrate, with the exception of the single pass. The single pass took just over 10 minutes. The strange thing was each file MainConcept file was around 70-80mb, whereas the single pass test was 126mb.
I then changed to Sony AVC which I understand uses CBR. I changed the bitrate to 8mps to try and comply with YouTube's guideline. This time it took around 8 minutes and 40 seconds to render the same one minute of footage. Checking the properties of the video the bitrate is more like 7.3mbps. The file size was 53mb.
I also tested the uncompressed AVI preset (producing a 6.3gb file) and DNxHD at 45mbps (I think the file was around 300mb) again each took between 8-9 minutes to render. So compressing these in handbrake would add to the time it takes to get a video ready for YouTube.
I was reading today that MainConcept is the preferred way to render for internet videos with current versions of Sony Vegas, and a couple of years ago Sony AVC was the preferred method.
Having looked at all the videos I can't actually see any difference between them watching them both on VLC and on the webplayer in Google Drive. Strangely I downloaded the SonyAVC video on my Chromebook, and the image is flickering (maybe a problem with their player? As it looks fine on Google Drive.)
I guess, I'm just wondering what other experiences people have had. If there is a different way to render videos a bit more quickly. Is 8 minutes per minute of video an acceptable time to render?
This is a complex topic. If you are doing your testes with FX's on the video it becomes even more complex. Here's why:
Rendering = Playback + Encoding
Since Vegas Pro 11, there has been playback GPU acceleration that is controlled under Options | Preferences | Video | GPU Acceleration. This is used for two things: (1) Playback of the media on the timeline, (2) GPU Acceleration of FX provided by Sony. This is based on OpenCL so video cards like the AMD Radeon HD and FirePro series will benefit greatly from this.
Each Encoder in Vegas Pro (i.e, Sony MXF, Sony AVC, MainConcept AVC, MainConcept MPEG2) may or may not support GPU acceleration of the encoder. Both Sony AVC and MainConcept AVC allow you to turn this on and off. Unlike the GPU acceleration for timeline playback, the encoders support both OpenCL and CUDA but only very old cards circa 2010. As you have pointed out, your GTX 960 is not supported.
To make things even more complicated, 3rd party FX like NeatVideo, Magic Bullet Looks, Boris BCC, have their own GPU support that is NOT turned off when you turn off GPU Acceleration under preferences and these plug-ins can and do use CUDA, OpenCL, OpenGL, etc. This is why we never add any FX when benchmarking render performance because this introduces too many variables.
I assume you have read my tests here on the COW in this thread here and here.
What I found was the same as you are seeing. Sony AVC is almost twice as fast as MainConcept AVC using CPU Only (i.e., no GPU acceleration of any kind) but that MainConcept AVC with a supported GPU was significantly faster than Sony AVC with a supported GPU by 3.6x!
It should be pointed out that since your GPU is not supported, and you are only using an Intel Core i5, you should expect to see very long render times and apparently... you are. If you want to improve render times, get an Intel Core i7 4 core or 6 core. It's not only about clock speed. The Core i7 architecture is a lot better for video ending and rendering. So it is common to see long render times with a Core i5 and no GPU support.
I agree with John for most of the points mentioned.
If YouTube is you workflow goal, there are other elements to consider as well. Here is how a video file moves through Vegas:
Source Codec -> Decompress and convert to RGB -> Vegas Engine (Timeline playback & Effects/titles) -> Encoder + Audio Encoder (HEVC,MPEG4,MPEG2,AAC,WAV,...)-> Write file to container(.AVI,.MP4,.MFX)
GPU assist (openCL) works at multiple points in the process:
Main Concept's encoder works on older cards with CUDA (<=580 series), it also works with OpenCL AMD cards (not sure where it stops, and not sure why it would stop with OpenCL support.) The MC encoder attempts to run more of the encoder on the GPU than say the Sony AVC encoder. With fast supported GPUs, this was almost an offloading of the processes the Vegas engine had to do. Remember these are slower system days of Vegas. There have been tests done that show the MC GPU encodes actually add artifacts to the image, vs encoding CPU only.
Sony's AVC is a bandwidth constrained encoder and not really a true CBR encoder. You can use "Bitrate Viewer" and "Media Info" to see this. XDCAM422-50 is an example of a CBR codec, this means anything encoded will write 50mbs stream if the scene motion requires it or not. Sony AVC uses the OpenCL combo (CPU OpenCL Compute Devices + GPU OpenCL Compute devices) and performs math that would return faster results than on the CPU alone. CPUs perform floating point math slower than the OpenCL combo. CPUs are rated in MFLOPS, where as GPUs are rated in TFLOPS, combining the CPU+GPU performance in the OpenCL device. So the Sony AVC encoder will use your OpenCL abilities on your system, but it may be that your CPU is so slow at keeping up you are not seeing much utilization. If you system is busy on some other aspect of the workflow, then the encoder will have to wait on frames to encode. System optimization to best support the entire workflow in Vegas will offer the fastest .MP4 encodes.
Optimizing your system to best support the Vegas Engine will get frame data to the encoder the fastest. This means optimizing Memory Size, Memory bandwidth (freq + timing), CPU bandwidth (i7 vs i5, # of memory channels supported), GPU performance (GFLOP performance, VRAM bandwidth, PCIe interface speed), Disk drive latency vs throughput (most drives are fast enough to deliver Prores,DNxHD,AVC compressed codec speeds.) CPU core count matters due not only to computational speed, but also the way Vegas and windows multitask the overall workflow.
Source codec can make a big difference in final encode speed. An intra frame codec may take more space, but LGOP codecs like AVCHD take more CPU computation to reassemble frames to be converted to RBG. Intra frame codecs have less system overhead that can be put towards other areas of the workflow. Sony codecs like XAVC-intra, HDCAM, XDCAM are better optimized source formats inside Vegas than .MOV(dnxhd,prores) formats, and in some ways even better than uncompressed/cineform.AVI. Remember Sony made Vegas to work Sony codecs the best coming from their cameras and VTRs. Support for other codecs should be looked at as in roads, and ways to export for delivery compatibility. Timeline footage in Sony.MXF codec will be the most edit optimized and stable as far as GPU support and effects.
OVERALL DATA QUANITY:
YouTube re-encodes all uploaded material to these standards - https://support.google.com/youtube/answer/1722171?hl=en
There is little reason to work uncompressed source material to be compressed to 8Mbs h.264. Some say upload h.264 files should be 10mbs-h.264, others say 2X the youtube bitrate, while I take a different approach.
Upload speed on your internet connection matters. My upload is fast, so I can upload large files in the same time someone else with restricted upload can upload an 8Mbs file. So I look at my workflow and find that rendering to XDCAM-EX(35mbs-VBR).mxf gets me both optimal quality and rendering speed due to optimizations in Vegas. YouTube supports converting XDCAM to the various MPEG-DASH frame sizes, and bit rates, and has no effective upload limit. Sure I am uploading more data, but who cares, set the upload and task switch.
60P is effectively doubling the amount of frame information to be processed. This means a slow machine will double the amount of time to encode a length of footage, vs a 30P file. Does your content really need 60P, if your system sucks and is slow? Are your viewers watching your frame rate, or the content of the material? Speed up your renders by dumping half the frames to be processed.
Hey, thanks for both of the replies.
I didn't come across those posts. Ive read them, and it makes sense. It's just one of those things I guess. It's a shame that modern Nvidia cards aren't supported. I guess it was the move to Nvenc. But I checked my card and it does support open cl 1.2, seems funny Nvidia doesn't optimise their drivers for this!
Sadly, I wont't be able to upgrade for a while - as i upgraded from an i3 4430, I don't even want to imagine the time it would take for that to render!
I do have one other option... I've still got my first pc I ever made... Ergh... Sure it has a dual core e8400 but it does have an ATI 5850, I wonder if that card would be able to do anything quicker even with the dual core cpu.
This is very detailed and perhaps a little over my head. But I do appreciate the explanation. I'll have to take some time to let it sink in (it's been a long day.) So Vegas is optimised for sony would transcoding my files to a sony format make much sense? I guess it would still be as broad as it is long.
As for 60p vs 30p I understand your point and would agree if it was live action. Personally I think watching a video of a video game with 60p does make a huge difference. I would argue that a 720p 60 video looks better than 1080p 30 - that's just my opinion. But you are right, if it was 60p live action, it has that wierd plasticy look.
I just rendered a 1 minute clip sony avc 720p 60 video (no effects) in 2 minutes. For continuity a 1080p 30 clip no effects in 1 minute 15 (suggested by John.) 1080p 60 again took 2 minutes, I would have thought the lower resolution of 720 might have sped it up a little. So my effects are adding 7 minutes render time per 1 minute...
I think its neat video that's killing the render time it's adding 6 minutes. all the other plug ins combined only add 1 minute to the render.
If that makes sense.
So compressing these in handbrake would add to the time it takes to get a video ready for YouTube.
Not if you use this. http://www.vegasvideo.de/vegas-2-handbrake-en Once you get it set up, it really is pretty slick. Rather than rendering to an intermediate, you are frame-serving directly to Handbrake. Within Handbrake you can also encode with Quick sync, which is a lot faster than CPU only. You can easily create a YT preset in Handbrake. The consensus seems to be that Handbrake produces better encodes, especially at lower bit rates. Since installing this, I have not used either MC or Sony AVC for encoding. Another option for better quality and faster encoding would be TMPGEnc's Mastering Works, which also works with frameserving.
You might also try the render to .MFX>XDCAM-ex-720-60p profile. This can be a lot quicker than AVC.
In the Sony AVC profile you are using keep the "Encode Mode" to auto.
The ATI 5850 is almost neck and neck with the Nv960. The ATI will have better OpenCL support, and has more OpenCL compute units than the NV card. The PCIe interface is slower on the 5850, but more compute units is better than the lesser NV960. For Vegas the ATI card would be a better choice.
NV960 - 2300 GFLOPS but lessor OpenCL support by NV driver.
ATI 5850 - 2000 GFLOPS but great OpenCL support by AMD driver. This card also has more memory bandwidth than the 960.
There could be other reasons for keeping the NV card, like other programs that take advantage of the card.
[Rob Clapham] "I think its neat video that's killing the render time it's adding 6 minutes"Oh yea! NeatVideo is a render hog. Forgot about that. This always slows down a render significantly. I use to use it up front. In other words, I would use it to clean up my video and then render that out and then use the cleaned up videos in my project. This way I don't pay every time I render the final project.
[Aaron Star] "...it also works with OpenCL AMD cards (not sure where it stops, and not sure why it would stop with OpenCL support.)"It stops at the Radeon HD 6xxx series because the programmers actually check for specific cards and disable GPU support if you card is not in the list. My Radeon HD 5870 works but my Radeon HD 7950 does not!
Thanks again for all the replies.
I think we're getting somewhere with the long render times!
I'll just try to summarise everything I've done so far and include you're opinions too. As John rightly pointed out, I really should have done my render test with no effects. Naively, my thought process was "these are the plugins I will use, so that's how I will test." Since my first post, albeit mainly testing with Sony AVC, as this seems to be the quickest option (without frame serving) I realise that my render time isn't as bad as it first appears.
The encoding preset I used was Sony AVC, I changed it so it was at 1080p 30 (29.970) and set the bitrate to 8mbps. With no effects, it took 1min 10sec to render 1 minute of video. This, to me is reasonable considering I don't have an i7, and can't use my Nvidia card. I'm not sure if you guys who specialise in video would find this unacceptable? But this is quicker than I expected, bearing in mind I had only ever used dual core processors in the past. I disabled all effects and NeatVideo was the first in the chain, I enabled NeatVideo and the render time jumped back up to just over 7 minutes. Considering my first post and the tests were 8 minutes with Sony AVC - All the other effects including adding noise to dither (which I guess wouldn't be needed if I didn't use NeatVideo?), Sony Convolution Kernel (sharpening), Magic Bullet Looks (cinestyle s-curve, shoulder tool), Sony Levels (computer RGB to studio RGB) added 1 minute of rendering time to my video. I did not test each of these effects to see exactly how much time each adds to the render - to me an extra minute to make my videos look a lot nicer is a reasonable trade off.
So as it stands NeatVideo is what is making my render times jump up significantly. As John points out it is a render hog. I did a couple more tests with NeatVideo, I added the preset back to my video (ctrl z - so it was the same noise profile in all my tests.) I went into NeatVideo's settings and you can change how it processes, so you can use CPU only, GPU only or CPU and GPU. CPU only gave me just over 7min 40sec (no other effects) render time. GPU only gave a render time of just over 6min 30sec. In NeatVideo there is the option to let it automatically optimise itself between CPU and GPU etc. I ran this tool, and it suggested to use 3 of my CPU's (i5 4690K) cores and all the memory of my GTX960 (4gb.) This render time then took 6min 10sec to render. So those with more powerful hardware, can possibly expect better results, if you weren't already aware of this feature. I will have to do a test later to see how it performs with all 4 cores of my CPU and GPU. I also wonder if NeatVideo is optimised for OpenCL, or if there are any other tricks to optimise the render time.
So to address everyone.
@John Rofrano You say to render NeatVideo upfront, do you have a preferred codec/encode setting?
@Wayne Waag I did see that in a post the other day. It sure does seem interesting, but also a little complex at first - I'm sure when it's set up, it's a piece of cake. I also wondered when reading the article if it's possible to do it 'manually.' What I mean is, is it possible to download debug mode's Frame Server only, and then click a few buttons to get it to work with Handbrake? Or is the more automated solution (with avisynth etc. and other downloads) suggested in the article the only answer? I only ask because I don't really like messing with my anti-virus, and I use Avast which was also mentioned in the article. Also Looking on debug mode's website under the usage instructions, they seem very concise and don't mention the steps provided in the article you posted. Which I guess is designed if you want everything automated? Please correct me if I'm wrong.
Also, I guess you could use intel quick sync with handbrake? Sadly I only have one monitor, but I do have it set up because I originally wanted to use quick sync with OBS, but the recording look hideous with very low bitrates (which you need to livestream.) If you're interested in quick sync, and have one monitor you can set it up with this tutorial
one extra thing I did, was to 'move' the extended monitor, so as to prevent the mouse moving to the virtual monitor. I'll post a picture later to demonstrate what I mean. Also, this workaround does work for OBS, it doesn't work in Sony Vegas to render with quick sync, you can get it to work with this tutorial
, but I don't have a spare cable to plug into my monitor at the moment - So I'm not sure if it works with Handbrake either. Also this workaround pretty much breaks the print screen function - so the snipping tool is your new best friend.
@Aaron Star I haven't tried the .MFX>XDCAM-ex-720-60p profile, I will be sure to have a look though and let you know if it's quicker. With the 5850 vs gtx960, I wouldn't really be able to swap the cards out. My i5/960 rig is mini-itx with a bitfenix prodigy case. You can fit a lot in this case! But, I only went with a 450w power supply. I think the main benefits with the 960, is despite the lower memory bandwidth, is that it is probably a lot faster than the 5850 for gaming. I believe that the 960 is slightly faster than a 760. My 960 also has 4gb, this is probably not a big deal - I only got this version because it was more or less the same price in the UK as the 2gb version. Ideally I would have liked a 970, but I didn't have the money at the time. I like the card for gaming, admittedly it's not the best but it's adequate. Another advntage - or so I thought at the time, was the ability to use Nvenc in OBS, which you mentioned in another post - sadly it looks even worse than quick sync, but that is at low bitrates to livestream. I'm sure if you increase the bitrate in OBS it would be more than good enough for local recording and an advantage over Nvidia's 'Shadow Play' I think I read that Sony Vegas couldn't read those files natively, but I'd have to check. So, if I was to use my 5850 it would be in my old computer with an overclocked dual core, I wondered if the GPU (5850) would be quicker despite the poorer CPU as the GPU is open cl.
I apologise for the really long post. Thanks again everyone for taking the trouble to post suggestions!
[Rob Clapham] "@John Rofrano You say to render NeatVideo upfront, do you have a preferred codec/encode setting? "When I did that I was using CineForm AVI files which are digital intermediaries that are designed for multiple renders without loosing quality. You could also use Sony MXF or just render back to the original format of the source file. One extra render is not going to show any visible artifacts as long as you keep the bit rate up.
I rendered my Neat Video plug in, in Sony MXF, the one with a default bitrate of 36mbps. Worked a treat and a slightly smaller file size than the original file from my camera.
I also sped up NeatVideo by changing the temporal radius down to 1. So it is now rendering at 11fps (or at least that's what NeatVideo kindly tells me.) I also have a question about dithering - I'll save that for a new thread though, as it may be helpful for others.
I haven't tried MXF outside of this, Aaron mentions that it is faster than Sony AVC - I wondered if YouTube accepts .mxf files? I'll have to try that out later. Without testing it, I can't say whether it is faster than Sony AVC, but between MainConcept and Sony AVC I can see very little difference in quality if any. But the render time with Sony AVC is a lot better on my current hardware - again I'll check Sony MXF tonight.
YouTube does accept xdcam-ex.mxf, XAVC.mxf, but not HDCAM.mxf ironically. HDCAM is older form of MPEG4 intra and was the go to format before Prores, but they support dnxhd and prores. Go figure.
I guess that's because dnxhd is in the Quicktime format? Maybe, I'm wrong. Seems strange though, as my experience with dnxhd, I haven't been able to change the bitrate (other than the presets) and I think the lowest is around 45mbps (from memory) which is way over YouTube's preferred bitrate.
I did get round to testing, whether .mxf was faster than the Sony AVC encode. Unfortunately on my machine the time is pretty much exactly the same. Not a huge deal, so I guess it's easier to use Sony AVC as that is ready for YouTube...
But the problem is the Sony AVC files I've tried out on YouTube, play back flawlessly on both my main computer, tablet and the Chromebook we've got kicking around. However on the TV (fire stick) the YouTube app, basically ruins the video with very out of sync audio.
Not sure if this is a problem, with Sony AVC or YouTube's TV app - as uploading the video, YouTube does warn that there 'may be' sync problems. But as I say, it works fine on multiple devices just not the TV. Strange.
I guess I'll have to try out MainConcept again, to see if that fixes uploaded videos on the TV. But, that will double my render time (compared to Sony AVC) - which I guess is the better price to pay, to ensure compatability.
I could actually use dnxhd, as when I tested it, the render time was comparable to Sony AVC (around 1:1 with no effects, 1:2 with effects) Sorry, if I got my ratio's wrong - hopefully they make sense. But I'll have to double check that.
I'll have to check out Wayne's suggestion for frame serving, as well.
I also wondered when reading the article if it's possible to do it 'manually.' What I mean is, is it possible to download debug mode's Frame Server only, and then click a few buttons to get it to work with Handbrake?
No. Handbrake will not directly read the "fake" avi that the debugmode frameserver produces. It is necessary to install the other software (Avisynth and Pismo). It may appear complex, but it only takes a few minutes to install and set up. If you install just the frameserver, you can render with TMPGEnc's Mastering Works.
In my view, installing and learning to use Debugmode Frameserver, Avisynth, and Virtualdub pays huge dividends in the long term since it opens up a vast array of filters and processes that are often better than the paid plug-ins and are free, a good example being Neat Video that can easily introduce the "plastic" look. Another example, fish-eye removal when using action cam footage from a GoPro.
Regarding your AV sync issue, this really sounds like a YT/Firestick problem. A question for you. If your goal is to watch "your" video productions on "your" TV, why would would you go to the bother of uploading to Youtube and suffer the inevitable loss of quality? It is very easy to stream directly from your pc or use an external storage device in conjunction with a dedicated media player or just a blu-ray player. My preference is actually a Sony blu-ray player (I use a 2013 model.) since it plays 1080 60P video files without issue and also supports the MKV container, which allows chapter selection. Just a few things to consider.
Thanks for the response. Good to know that Handbrake can't read the "fake" avi files on it's own - and that the article you originally posted is the way to do it properly.
I think removing the GoPro's fisheye is done with VirtualDub isn't it? I have used VirtualDub ages ago, for the stabilisation plug in - very useful stuff. I agree that NeatVideo can make things look very plasticy - I've had much more success with it now though, by altering the settings (something for some reason, I never thought to do, until I encountered that plastic look.) Since then I reduced the temporal radius from 2 to 1, which sped things up a lot - and in my opinion didn't make the noise profile look much 'worse.' Even reducing the amount of noise being removed to -60%, the plugin managed to remove a lot of noise and not look like plastic. I really should invest in some better lighting though, to get my ISO down.
Well, hopefully the videos I put on YouTube will be seen by a 'wider audience' that is if people want to watch them. So, it isn't for just for me to watch - if that makes sense. I agree, if it was a private/home movie or something similar uploading to YouTube isn't the best approach. I just put a couple of videos up as I was testing out Acon Digital's "Deverberate" plug in. I had already played it back on headphones and it sounded good to me, and wanted something else to compare it to other than laptop speakers. That's how I encountered the audio and video sync problems with the Firestick. I did think it was odd though as the Sony AVC files I uploaded, mentioned during the YouTube upload process that video and audio might be out of sync. But as I already said, it was fine on everything I tested except the Firestick.
I wonder if the message I saw was relating to watching videos on the TV, whether it's a Firestick, Roku or something similar? Come to think of it, a test I uploaded last night had the same problem and that was encoded with Handbrake (after I made a Sony AVC video) - as I wanted to see if the Audio/Video sync issues still remained, and they did - So I guess you're right and that it is a problem with the Firestick. I'll have to watch some other videos people have uploaded to see if it's out of sync there (I mainly use YouTube to 'watch' radio type broadcasts with one picture, so maybe that's why I haven't noticed it before.) The YouTube message saying that there may be problems, also directed me to their help page on how to encode a video. But the page was only talking about Apple NLE's and how to use QuickTime. The videos I've uploaded on YouTube are private, and have been up a couple of days so it's not that YouTube hasn't processed them yet.
Sorry, I couldn't figure out how to edit my above post.
Yeah, it's definitely to do with YouTube on my Friestick (I wonder if other streaming devices have similar problems?) - played a video and it was just as out of sync as mine were.
Still, I think it's strange that YouTube said I may experience sync issues. Does this occur when other people upload their videos?
Sorry, I couldn't figure out how to edit my above post.
It's been changed. To edit, simply hit the reply button to your post and to the left you will see the edit button.
I think removing the GoPro's fisheye is done with VirtualDub isn't it? I use an Avisynth script that you simply open in Virtualdub. I've put together a Vegas script for automatically going directly from events on the Vegas timeline to the avisynth script in Virtudaldub. The "defished" file is then added as a Take to the original event. Just select the events, and walk away till the processing is done.
I agree that NeatVideo can make things look very plasticy
A far better denoiser IMHO are some avisynth scripts written by John Meyer. Whereas Neatvideo is mostly a spatial filter, his scripts are temporal in nature. They are a lot more difficult to run, but the results are better and they don't cost anything. I find that with NeatImage for stills, you often get that plastic look as well. It's very easy to overdo the amount of noise reduction.
So, it isn't for just for me to watch - if that makes sense.
Makes perfect sense. If it's for internet distribution, you really don't have much choice. As far as the potential out of sync warning, I get that most of the time as well, even when there is no dialog. Also the message that your video is shaky. Lets face it. Lots (most?) of people will watch YT stuff on their phone or a tablet or pc that's got a slow internet connection. Even with my 100 MBps download, lots of videos are pretty shaky and not very smooth. It's all to easy too forget that content is still king.
I still think NeatVideo can be pretty effective, once the settings are optimised. But saying that, I haven't had the benefit of comparing it to the denoiser you mention by John Meyer. So it's more than likely a big improvement - I'm interested in checking this out, but I don't have an i7 and if it's quite intensive, it might be better for me to stick with NeatVideo until I can upgrade my computer.
Your post has opened my eyes a lot. I think I've been obsessing over trying to get everything to run quickly etc. That I've forgotten to actually make the videos because I enjoy it!
You're absolutely right by saying "content is king." The amount of stuff I watch, sometimes with poor audio, sometimes poor video, sometimes both - and the reason I carry on watching? Content. Sometimes, you need someone to point the obvious out!