program to optimize for youtube

[intracube]

Another finding (at least it's new to me; I don't use YT that much), is that 720p versions get allocated more audio bitrate than their SD counterparts. It used to be the same as far as I can recall. A 480p version might get ~128kbps but 720p gets ~150kbps .

I think YouTube have done that for a long time. See the table here:
http://en.wikipedia.org/wiki/Youtube#Quality_and_codecs
*not sure how up to date it is

Another bug I noticed back in September was with media files with no audio track. After upload, the picture would just be grey. Don't know if that's been fixed yet. Also, it might have been specific to the container/codec I was using.

[intracube]

I've uploaded a new test which works properly

Out of curiosity, what program do you use to create those test videos?
Thanks.

A mixture of Blender, GIMP, audacity & ffmpeg.

[Knightmessenger]

Youtube started unlocking the time barrier for popular accounts that had no infringement claims or violations back in 2010. Eventually they started allowing more and more accounts unlimited time so long as you have some videos up and in good standing.

But does anybody know why the video is jerky. Even if youtube does throw out one of the fields, that doesn't explain why my previous uploads all had smooth motion. (despite more noticeably compressed and more artifacts.)

I'm wondering if I should try again without the denoise filter as maybe trying to blend all these frames when the camera wasn't steady didn't go well. Or would the output file size matter? I set the target size to be 300mb, would 500 mb change anything about the jitteryness of the frame rate?

[intracube]

But does anybody know why the video is jerky. Even if youtube does throw out one of the fields, that doesn't explain why my previous uploads all had smooth motion. (despite more noticeably compressed and more artifacts.)

The main difference with the video you linked to in post #11 is it was shot at the telephoto end of the zoom range to pick up the people on the stage, whereas the videos in post #5 were shot wide angle.

Shooting hand-held while zoomed in is going to greatly magnify any vibrations.

Also, it looks like all your videos have been shot with a fast shutter which is making the jitter even more noticable (very little motion blur) - maybe ~1/120sec instead of 1/60 or 1/59.97sec - which would be more typical for video.

[intracube]

Another YouTube oddity; YouTube delivers two of your videos with the standard Flash player, but this video uses HTML5/webm (which skips/stutters on my computer).

I normally have webm disabled and get the message "Your browser does not currently recognise any of the video formats available.".

[zoobie]

Hand-held is also the difference between rank amateur and semi-pro/pro.
But it sure is convenient, isn't it? Just what the typical consumer wants.
The video is a result of exactly how much effort you put into recording it.
I lug around a 35 pound tripod for a camera less than a pound.
You can imagine my results...

[Knightmessenger]

no, the stuttering is definitely encoding. I have the avi file and the frame rate and motion is the same with the original files. (yes I know zoomed in magnifies jitter, yes I forgot my tripod and yeah holding the camera above my head over people in the audience is very exhausting)
At this point I don't care that the video isn't the greatest production. I've learned from it and I have other things I'd like to move onto that I think turned out much better. But it can't have any noticeable errors that was not present in the original video file.

[intracube]

no, the stuttering is definitely encoding. I have the avi file and the frame rate and motion is the same with the original files.

If you've put all the videos through the same conversion process, then unless there's an obscure bug with the software any differences are to do with the source files.

I've downloaded this video, gone through it a frame at a time and can't see any sign of irregular frame duplication or frame dropping (other than the 59.94->29.97 conversion).

I don't think the encoding process is introducing stuttering for that video alone. It's just that that particular video needs 59.94fps to avoid looking jittery because of the amount and type of movement.

Also, It's not just the amount that the camera is moving, but the type of movement - a high frequency jitter rather than a swaying movement. What camcorder was the footage shot on? I'm guessing a small, light camcorder with very little mass.

zoobie is right that pros will usually use a tripod. But even when they don't, professional cameras are relatively heavy and shoulder mounted. The added mass of the camera pretty much eliminates high frequency jitter. I used to use a shoulder mounted S-VHS camcorder and loved it for the steady video it produced - it had no form of lens/image stabilization either.

As poisondeathray has said, your best bet would be to use software stabilization to make the video more presentable when it's converted to 29.97fps.

[Knightmessenger]

Hi8 TRV 615, I never liked the really tiny cameras either as they are so light, it's harder to hold them steady. I went ahead and uploaded the video with different encodes.
decomb and denoise
http://youtu.be/32qQVvtkt0s

decomb only
http://youtu.be/5C3vBZaC-kQ

no decomb, no deinterlace, no noise reduction
http://youtu.be/qP3hSPNAVPE

deinterlaced on "slower" but no decomb
http://youtu.be/Du0_Ot2jelQ

The interlaced one has the smoothest motion but the jaggies are really obvious. I do think the one without noise reduction looks slightly better than the first.
What do you think?

Also, if I was to upscale it to a higher resolution for a higher bitrate, what size should I pick? This is 640x480 which is aspect ratio 4:3.

[intracube]

The interlaced one has the smoothest motion but the jaggies are really obvious.

I can sort of see what you mean. Because of the fast shutter and lack of motion blur, showing both fields softens the edges. But the jaggies are too objectionable IMO. I would go with the second video (decomb only), or try image stabilisation software which could give a big improvement. Take a look at this:
http://www.youtube.com/watch?v=9i1iJgesPbQ
EDIT: it looks like that video has had a bad framerate conversion on top of the camera shake. If you're just going from 59.94->29.97 you shouldn't see that effect at all.

I do think the one without noise reduction looks slightly better than the first.
What do you think?

I prefer the one without noise reduction. It preserves a bit more detail.

Also, if I was to upscale it to a higher resolution for a higher bitrate, what size should I pick? This is 640x480 which is aspect ratio 4:3.

960x720 is the next step up in frame size - the central 4:3 area within a 1280x720 16:9 frame.

[Knightmessenger]

EDIT: it looks like that video has had a bad framerate conversion on top of the camera shake. If you're just going from 59.94->29.97 you shouldn't see that effect at all.

you mean the interlaced one? I can tell you that that file looks way noticeably interlaced in WMP than the original file which is also interlaced. With no video filters such as de interlace applied, vid coder somehow screwed it up. Is there a better software out there besides vid coder?

note that I added the link to the last one. I think that looks the smoothest and doesn't have jaggies but it does appear to have one of the fields discarded rather than interpolated as it looks less detailed when you pause it than the others. (the labels on the water bottles in the close up of the MC's in the first clip)

[poisondeathray]

As poisondeathray has said, your best bet would be to use software stabilization to make the video more presentable when it's converted to 29.97fps.

This


http://www.youtube.com/watch?v=gvwi0wtpSmY&hd=1


Because the amount of shake is so bad in the stage video, some frames are completely distorted and out of focus . But you can interpolate / reconstruct some of them using mvtools, but the side effects can be warping distortions in some frames .

Is there a better software out there besides vid coder?

Yes there is better software, but not "1 click" sort of thing. Vidcoder is fine for encoding, but it doesn't do a single thing on that list posted earlier for optimizing encodes. Yadif deinterlacing is overrated. Speed is great and it's fine for real time deinterlacing, but it leaves substantial artifacts and produces lower resolution. Have a look at the comparison video posted earlier. I can post more examples if you want

In case you missed the "list":

Depends on what your definition of "optimize" is

When people say "optimize for youtube", that usually includes higher quality deinterlacing, stabilization, specific denoising , levels corrections, high quality upscaling without aliasing, among other things. No "easy to use" program will offer all of these. There is going to be some trade offs involved here

On that last video, you can fiddle around with deinterlacing, denoising all you want, but it doesn't address the real problem in that specific video, which is the lack of stabilization . On your first video (with boba fett) , those other factors on the list will make a more significant difference to the image quality, because the amount of shake is less and not so overpowering.


Your videos have black levels crushed (your're thowing away shadow details, but there is recoverable data there). See the attached boba fett screenshots. In the waveform tracing, the areas in brown won't be displayed on youtube or most players and devices - notice how much data is in the brown area. Your stage video is like this too


Software stabilization has a list of negatives or side effects. In addition to processing times, you get large black borders as the software attempts to re-center the frame. So you have to choose between either zooming into the image or various border fill options . So, as mentioned earlier , it's much much better to stabilize while shooting.

[intracube]

Note: I thought I replied to this thread last night. But I can't see any evidence of the post now. Weird.

EDIT: it looks like that video has had a bad framerate conversion on top of the camera shake. If you're just going from 59.94->29.97 you shouldn't see that effect at all.

you mean the interlaced one?

No, I mean the video showing the image stabilisation in After Effects.

I can tell you that that file looks way noticeably interlaced in WMP than the original file which is also interlaced. With no video filters such as de interlace applied, vid coder somehow screwed it up. Is there a better software out there besides vid coder?

I can't suggest specific software as I'm a Linux guy I usually use ffmpeg to convert videos - there's a version for Windows if you want to give it a try. But be warned, it's a command line program.

note that I added the link to the last one. I think that looks the smoothest and doesn't have jaggies but it does appear to have one of the fields discarded rather than interpolated as it looks less detailed when you pause it than the others. (the labels on the water bottles in the close up of the MC's in the first clip)

Both the videos look very similar to me. The variability of the compression artefacts swamps the differences between the de-interlacing filters. Each video looks better than the other - depending on exactly which part of the video is compared. But all things being equal, using a de-interlacing filter that uses both fields (like yadif) will give better results than just dropping a field.

Regarding the encoding options, I'd choose "constant quality" and try different settings on a short clip to determine the best compromise between bitrate/quality. Aim for 15MB/min (2Mbit/sec) as a good starting point for 640x480 video destined for YouTube. Use h264/x264 video codec.

[intracube]

As poisondeathray has said, your best bet would be to use software stabilization to make the video more presentable when it's converted to 29.97fps.

This
http://www.youtube.com/watch?v=gvwi0wtpSmY&hd=1

Very nice job pdr

Because the amount of shake is so bad in the stage video, some frames are completely distorted and out of focus .

That's because of the motion blur, right?

This has got me thinking about the post you made a few weeks ago. If that photoshop algorithm could be applied to video after it's been stabilised...

[poisondeathray]

As poisondeathray has said, your best bet would be to use software stabilization to make the video more presentable when it's converted to 29.97fps.

This
http://www.youtube.com/watch?v=gvwi0wtpSmY&hd=1

Very nice job pdr

Thanks, but it's nothing special, just common sense approach ... fix the worst thing first

There are free programs for stabilization . I used deshaker in vdub here, and I think there are a few for linux too. You don't need After Effects for this . The one saving grace of many older DV camera is that they are CCD , so no CMOS rolling shutter type artifacts and introduce a whole set of other problems

Because the amount of shake is so bad in the stage video, some frames are completely distorted and out of focus .

That's because of the motion blur, right?

I think it's a combination of auto focus and shutter angle. Because of the quick camera jerks, it "auto" mechanism tries to re-focus so every few frames is almost unusable, even if you stabilize it

This has got me thinking about the post you made a few weeks ago. If that photoshop algorithm could be applied to video after it's been
stabilised...

I haven't tested the beta, but most deconvolution algorithms used for directional blur work on calculating directional vectors e.g. focus magic , topaz infocus . In reality , they don't work that well, or only under cherry picked situations. I'm very curious to see how the Adobe approach pans out

[Wissen]

I've posted about this before but what I want to know this time is if there is an easy download of a program that lets you optimize video files for youtube. I only have Vegas version 4 so it's not the most up to date.

I need something similar to Compressor on a Mac (I have Windows) that can de-interlace, undo the stretch (720x480 -> 640x480) and basically make the file a lot smaller. Upscaling to 720p would be a nice feature since I've been told many times here that enables more bitrate from youtube but it's not absolutely necessary.

Is there a basic easy to use software I could download to do this?

Have you considered simply upgrading to Vegas 11? It does excellent YouTube optimization, the best I have seen.

[Knightmessenger]

I think it's a combination of auto focus and shutter angle. Because of the quick camera jerks, it "auto" mechanism tries to re-focus so every few frames is almost unusable, even if you stabilize it

As of fall 2008, I always use manual focus. Can't believe I thought auto focus was acceptable for so long. Manual is especially good for something on a stage like this because whenever somebody in the crowd stands up, the camera would brighten and blur up on the auto settings.
It's one of the reasons why I've stuck with my Hi8 for all these years as most newer non professional cameras don't even have the option. With the crushed blacks, I used manual exposure and I simply set it too low. I overcompensated for the fear of over exposure and should have realized the LCD screen is not nearly as accurate as the black and white viewfinder.

I'm very cautious about experimenting with any stabilization because I know that can only take out frames and try to guess what the replacement ones would look like. I imagine it could easily introduce a lot of smearing or ghosted frames similar to what happened with the DVNR on the THX version of the Star Wars Trilogy. (the same transfer that was used for the original versions on dvd)

I do intend to ask for the latest version of Vegas for Christmas.