Coldstreams Online Video

This past year as seen the phenomenal growth of the Digital SLR (DSLR) market for cameras that now also shoot HD video – either 1280 x 720 or 1920 x 1080, depending on the camera. This market is about to take a huge leap forward over the next few months as all the camera makers introduce new products that begin to support video needs much better than their first or second generation cameras.

DSLRs are becoming popular for video photography because:

1. Many still photographers have quite a collection of lenses that are a much larger investment than the camera body! Now they can put those to use for video too.
2. Having the combined function of both high quality still and video in one camera is very convenient. In a national park in Canada recently, I was carrying a non-video DSLR, an extra lens, extra battery, a Canon HV30 video camera, a wide angle lens, extra tapes and extra batteries – plus a video tripod – on an 8.5 mile challenging hike. That hike convinced me that next year, I’d like to have a video capable DSLR! Less stuff to carry!
3. DSLRs are capable of narrow depth of field in a way that most video cameras cannot do – except for extremely expensive professional cameras.
4. DSLRs can use special effect lenses to create unusual optical effects, such as making the real world look like miniature models! (My wife thinks that is funny since so many in the movie making world work very hard to make their models look like the real world – now we try to make the real look like a model!) For example, see this.
5. DSLRs can be used less obtrusively to shoot video than using larger cameras like the XH A1 or Sony EX-3.

To video photographers, DSLRs have their drawbacks – so far. Typically their audio feature set is limited.  They display video only on exterior LCD panels, which may not be tiltable (this is changing for the new models) and which is hard to see in daylight. They limit maximum clip length to 12 minutes or similar. They often do not auto focus in video mode and there is no such thing as a motor driven slow zoom – they are all manually operated.

But the one feature that stands out in videos shot on DSLRs is their depth of field capabilities. They can achieve narrow depth of field because their image sensors are much larger than typical video cameras. The larger the sensor size relative to the lens, the easier it is to have that narrow focus range.

The good news is that you can kinda sorta achieve some DoF capabilities with the XH A1. At the wide angle setting the lens opens up to f1.8. It stops down a bit on telephoto. If you manually control the aperture (set to the Av setting) you can achieve some nice DoF effects. Its not the same as a DSLR, but there are some decent views that you can obtain this way. I’ve done some experiments to get some effects I wanted – and its okay.

A possible problem is when you open up to f1.8 in bright sunlight, the shutter speed may drop to 1/600 or worse. For moving subjects or pans, this may create too much of a strobe effect at 24f or 30f frames per second.

An alternative is to use neutral density filters. I have not yet tried this – but with external ND filters (in addition to those built in to the camera) it should be possible to shoot at wide apertures with a lower shutter speed like 1/30th to 1/100th of a second.

Anyone try this yet?

I previously played my own HD videos (videos that I created) at 1920×1080 by streaming them off a slightly older Windows XP PC over a network to an Xbox 360 connected to an HDTV.

Two things happened that killed that solution:

1. The Xbox 360 eventually died completely (already been down the “red ring of death” problem in the past).
2. My youngest child is now old enough that she did not care if we replaced the 360.

Looking for a low cost solution I re-assigned my old 3.06 Ghz single processor machine still running XP to play videos through the VGA connector on the HDTV. I still used that machine for some software development but spent today transferring over the last files and programs I needed on to my much newer multi-core desktop that I use for development and video editing.

The good news is that this all went well.

The bad news is that any time I tried to play a 1920×1080 MPEG4 video (most of mine are encoded at 4 to 8 Mbps depending on the content), the images stuttered badly. One in particular looked like I was watching a slide show, not a video!

I tried using Quicktime, VLC and the Media Player Classic Home Cinema software and all had various kinds of problems from stuttering to false colors (VLC).

Looking around I discovered that the problem is due to H.264 decoders that are too slow and can not keep up with full frame video.

But there is a solution – CoreAVC 2.0 and CoreAAC codecs. They have optimized their H.264 decoders for far faster performance than all the others.

Once I bought and installed those codecs, I can now play the 1920×1080 H.264 videos in Windows Media Player and get smooth video. No frame drops that I could see. Plus, the fan on the PC only kicked up half way to keep the CPU cool – so the fan noise was lower too.

For the $18 cost of the software, I’ve replaced the parts of the Xbox 360 that we really wanted – streaming our own videos. Plus we can now watch YouTube, Hulu, Vimeo and eventually Netflix videos too. And I’m going to have give Flight Simulator or X-Plane a whirl soon too

The CoreAVC and AAC codecs are working well for me. You might take a look at them if you are having trouble playing 1920×1080 videos on an older computer.

For a software project, I had to recently update my old Windows XP machine to a newer machine running Windows 7.

I decided to also look at the free demo version of Sony Vegas Movie Studio Platinum 10 editing software.

Update: The following review is not quite correct. My original test footage was inadequate to test this out. Here’s the scoop: Vegas Movie Studio Platinum 10 will work with true 24P video. You can set a timeline to 23.976 and import video clips that are true 23.976. However, video recorded in the 2:3:3:2 24p “pulldown” modem which involves mapping 24p images into a 60i interleaved stream, will not work unless you use a separate program like Voltaic HD or CineForm’s NeoScene to remove the pulldown prior to editing.

I still really like Sony Vegas Movie Studio Platinum 10 – in fact I have ordered a copy – but it does not do 2:3:3:2 pulldown removal. But once I had everything in some form of 24p – including a combination of HDV, SD and AVCHD transcoded to remove the pulldown – I was able to edit the entire mix just fine in Vegas.

For a test, I gave it what I thought would be a difficult challenge – I combined 24PF video from an Canon HG10 in AVCHD format, with 24PF from a Canon HV30 in HDV format, with a 24P DV video stream. For those not familiar with those terms, 24PF is a 24 frames per second video stream embedded into a conventional 60i (interleaved) set of frames. To edit a 24PF video requires removing extra frames used to fit the 24 full frames into a 60 half frame (60i) video stream used by conventional video recording.

Incredibly – it just worked!

I set the Project Properties to 23.97 IVTC (inverse telecine) option. I then imported the media – in the case of the AVCHD video clips, these were already on the system hard disk. The HDV was imported from tape using the Sony Vegas video capture function. The 24P DV video was actually shot as 60i on a Panasonic DV camera, imported and converted to 24p using DVFilm Maker.

And it just worked! I didn’t have to do any futzing around – it just worked.

I output the result to an MP4 file at 1440×1080 and the results are excellent.

Since the DV video was shot outdoors at an evening concert, the colors were muted. I used some Vegas filters to slightly increase the contrast and saturation, and added a bit of sharpening. This was then scaled up to the AVCHD/HDV frame size, producing an excellent result.

I’m still using the demo/trial version but I think Sony Vegas Movie Studio HD Platinum 10 is a keeper. I’m pleased so far.

You can learn more and download the demo version at Sony Creative Software.

Each year, Spokane, Wa holds an annual Lilac Festival and Armed Forces Day Parade. This is the largest evening Armed Forces Day Parade in the country, and one of the larger parades of any type on the west coast.

This year, 211 units participated in the parade, with 46 high school marching bands from throughout the state (and some from out of state). The parade typically has around 50 + or – bands. If you like marching bands, this is a neat parade!

Over the years, I have followed the West Valley High School Eagle Marching Band since my own kids previously went through that band program. While I no longer have any of my own kids in the band, I still run their web sites and some IT functions behind the scenes, and do some media work for them, all as a volunteer.

Video taping a marching band in a night time parade is difficult. If you stand on the curb and shoot video, the band goes by once and its all over with in 20 seconds.

The only solution is to march with the band. But that adds a lot of challenges, even more so at night. I cannot use auto exposure in the camera since the camera tries to add 36 db of gain to make the night time look like daylight and covering everything in awful noisy graininess. Similarly, audio levels can not use the ALC to adjust audio level automatically – that makes the entire band and crowd noise come across as loud mess.

The opening scene of the following video is shot near the TV coverage area. I set the gain to 0 db and shot at 1/60th of a second in 30p mode, Tv exposure mode, on the Canon XH A1. Once the band passes the TV lighting, I manually switch to 12 db gain and the XH A1 has sufficient aperture to automatically set the f-stop the rest of the parade.

Why not use 1/30th of a second rather than 1/60th, considering shooting at night? Two reasons. One, 1/30th of a second produces some blurring, but does enable using +6 db gain instead of +12 db gain. But the second reason is that motion stabilizing software (see below) does not work well with blurred images. Sharp images work much better and that requires at least the 1/60th of a second shutter speed (based on some experiments I did).

I also use both a short shot gun and the on camera mic during the parade. For the long shot at the beginning, I used the shot gun mic to cut down the crowd noise. But once I start moving with the band, I switch to the internal stereo mic to get better coverage (less directivity) of the sound pattern of the entire band.

At about 1 minute 35 seconds into the video, you’ll see my fake “Steadicam” in action. You’ll see this type of shot throughout the entire video.

How do I do this?
1. I use a monopod and attach a 3 pound weight to the bottom of the monopod. This creates a simple dynamic stabilizer. Imagine holding a stick with a weight at the bottom. As you move the stick around, gravity naturally pulls the weight back towards the center. The amount of weight and the length of the monopod influence the dampening effect. I use only a 3 pound weight because I have to carry the camera gear for several miles (the parade itself is 3 miles long, plus I usually walk 2 or 3 additional miles before and after the event).

2. Learn how to walk smoothly. Watch marching band performers – they “roll step” as they walk. This means rolling your feet in a smooth motion rather than stepping, greatly reducing your vertical motion. Surprisingly, you can learn to walk backwards for an even smoother effect which has to do with how your body mechanics works out. Add a slight crouch and slightly bent knees and with practice, this technique combined with the weighted monopod can product surprisingly smooth walking shots.

3. In post, use motion stabilizing software. There are several applications available today that do a good job of analyzing images for motion and then shift your images around to create the appearance of stability. Since they has the effect of adding black edges to your video after shifting, most also slightly enlarge your original video to eliminate the black edges. For this video, I just used the built in motion stabilizing feature in Apple’s iMovie 8. It took about 3 hours to analyze the HD video I had shot.

Even with motion stabilizing software, you will achieve the best results if you also use the tricks outlined in (1) and (2). You can see the effects of this fake “Steadicam” throughout the videos, but if you are in a hurry, you can see it in the first scene in Part 2, below.

Other tricks I use during the parade are to use the monopod for overhead shots, including walking shots. Sometimes I take a video clip standing in the midst of the band but did not do that this year. It would also be nice to take some low level shots from near ground level, but there is never enough time.

By the end of the parade, I am drenched in sweat and exhausted. In order to get many shots of both the front and back of the band, I have to shoot while the band marches by, then run back up to the front of the band.

This was shot using the Canon XH A1. I have shot daytime parades using an HV30, external BeachTek audio mixer, and a short shot gun XLR mic, plus wide angle lens (always go wide angle for these things). I use the XH A1 for the night parade because of its better light gathering capacity of the big lens and larger imagers. But it does weigh a bit more!

Part 1 of 2:

Part 2 of 2:

This afternoon I downloaded and tried out the Sony Vegas Movie Studio 9.0b trial version. Vegas is unique in that it is able to edit AVCHD directly, without conversion.

But as I learned, you’ll probably need a dual core processor.

The process was slow on my 3.2 ghz single CPU PC with 1.5 GB of RAM and running Windows XP.  Vegas quickly imported the clips from the camera, and arranging them on the time line was quick and easy. BUT – all videos stuttered when I tried to play them without using the “render to preview” step.

By the time I started doing rendering so I could watch in real time, the whole process had bogged down. Converting to other formats was not real fast on the 3.2 Ghz machine, either.

Seems to me that if your system really can not handle true AVCHD editing, then you are best off doing a conversion to another editable format.

On the Mac, I’m likely to stick with Voltaic HD 2.x for both transcoding and elimating the 24p pulldown frames. Voltaic HD is also available for Windows. I have not tried the Windows version yet, but the idea is the same – it should transcode from AVCHD in to (probably) an .AVI file that you can use in your favorite HD capable video editor. (Note – not all video editors can handle 24p – you’ve been warned.

Now that Voltaic HD 2.02 properly converts AVCHD to AIC and removes 24p pulldown frames – I will probably just use that. Sure, there is a transcoding step from AVCHD to the AIC format. But it is darned hard to see any meaningful loss except for a very slight softening of the colors. Since my destination is eventually to MPEG4 files that I play on my HDTV any image losses in the AIC transcoding are irrelevant.

More info on AVCHD and also 24p is available in other posts on this web site. In other posts, I describe how you can use iMovie and the free program JES Deinterlacer to process hv20/hv30 24p, or AVCHD with 24p video frames.

For now, I’m likely to use Voltaic. It is not real fast but it does produce the best results for when you want to be picky about images.

I wrote in the past about dealing with AVCHD and 24p editing on the Mac OS X.

I just installed the latest version of HandBrake, the free video converter for Mac and it easily converted AVCHD directly to mpeg4 video files. Unfortunately, it does not know what to do with the 2:3 pulldown removal required for the 24p on my HG10. Still, if you’ve got 30p or 60i video in AVCHD and want to go direct to MPEG4, this would be a good solution.

Unfortunately, Quicktime doesn’t know what to do with the resulting mpeg4 file, but other programs, like VLC and MPEG Streamclip, played the mpeg4 file just fine. Go figure.

Update: I have been experimenting with Voltaic HD 2.0.2, which now correctly processes the 24p AVCHD files from my Canon HG10. I have successfully transcoded from AVCHD with 24p pulldown removal to AIC, Photo JPEG and uncompressed 4:2:2 and more.

• Uncompressed 4:2:2, not surprisingly, gives the best result. But an 82 Mb 40 second input file becomes a 3.7 GB uncompressed file! The most noticeable change is enhanced saturated color in the 4:2:2 color scheme versus AIC and Photo JPEG – and some improvements in subtle color graduations that leave compression artificacts in AIC and Photo JPEG. But the huge file size makes this unusable.
• It is very hard to tell much difference between AIC and Photo JPEG. They can pretty much be considered equivalent. The Photo JPEG codec definitely softened the colors while the AIC codec seemed to have a slight softening.  The quality, otherwise, is basically identical – however, the Photo JPEG version is about 1/3d smaller than the AIC file.
• I  downloaded and used the free Avid DVxHD code. Transcoding to this codec produced results similar to uncompressed but with an 800 MB output file instead of 3.7 GB. Still, that is a ten times file expansion. I also used the AVID DV100 codec which produced a 446 MB file with better color than AIC but the compression issues seemed no better. The Avid Meridien codec produced slightly larger – but all AVID codecs had better color.
• Conversion to HDV did not work, even though the original AVCHD is in 1440×1080. This appears to be a bug in Voltaic HD and this is supposedly a feature of Voltaic.
• I also have an XDCAM 35 Mbps variable bit rate codec (1440×1080/24p) that I believe came with FCP version 5.x. Unfortunately, like HDV, this did not produce a proper output file. If it had, the file size would have been smaller than AIC and provided better compression and color than AIC.
• The

At this point, for working with AVCHD/24p files that require pulldown removal, the highest quality will come from using Voltaic HD to transcode and do pulldown removal in one step (versus 2-transcode steps required in the iMovie–>AIC –> JES Deinterlace to inverse telecine –> AIC). As to which codec to use? I’ll probably stick to AIC.

Video News » Blog Archive » YouTube Adds Automatic Subtitles.

The new feature can automatically generate on-screen captions, which is very useful and valuable to those who are hearing impaired. The system can even translate to other languages.

As the narrator in the linked video notes,  “sometimes the automatic captions are pretty good”

Seriously – this is a fantastic new feature. Good job, YouTube!

There are mini-disk products that you can connect directly to your Firewire equipped camcorder to record direct to disk. But then tend to range in price from about $600 to $1800 depending on used vs new, capacity and vendor.

An alternative is to record direct to your notebook computer.

If you have a Macbook or Macbook Pro, a very easy way to record direct to disk is to connect the Firewire output of your camcorder to the notebook computer and then run iMovie and import from the camera.

The camera should be in its “camera” mode but does not need to be recording to tape. Whatever the camera sees will be recorded direct to disk.

This works with my Canon HV30 in HD mode and I’ve done it using both iMovie 8 and the older iMovie HD (version 6) connected to a Macbook.

On Windows, there are several software utilities available that will enable you to do the same thing. While I am typing this on a Windows desktop, I do not have a Windows notebook on which to test this out!

The main advantage to doing this is to overcome the occasional tape dropout problem that tends to plague HDV format.By recording direct to your notebook computer disk, who cares about tape dropouts! (Caution – you may want to use a longer Firewire cable to keep the notebook away from your camera mic, especially if the fan kicks on to keep the CPU cool. Not all cameras have sufficient drive signal to use a longer cable, though. So be sure to test out your configuration first!)

On SD recordings, a video dropout typically lost a single 1/30th of a second frame. If you even noticed, you could always copy an adjacent frame and no one would notice.

With HDV you can lose up to 1/2 second per dropout – and I guarantee, everyone will notice!

Two other steps to avoiding dropouts are to clean your video heads in the camera every 5 to 10 hours of recording – I use a Canon cleaning tape for about 10 seconds but I’m told most any cleaning tape is fine. The other important step is tape quality – I used to use TDK tape all the time on my SD camera with excellent results – but the SD tapes always had dropouts when recording HDV on standard TDK tape.

I switched to Panasonic AMQ (HDVM63AMQ) tapes and have now recorded probably 75 hours with excellent results on that tape. I buy mine from TapeStockOnline.com. They have consistently quick order fulfillment and decent prices. If you are used to buying standard miniDV tapes at the local discount store  you’ll find that high quality tapes for HDV are more expensive – currently $5.25 in a minimum order of 10 units.

There are other brands that cost both more or less than these but I’m sticking with what has worked well for me. Hopefully this note provides some idea for you to try if you are plagued by video dropouts!

I’ve seen some confusion on the online forums regarding shooting in 30p (or 24p) and then importing into a video editor.

Why does my 30p video show up as 60i in <name your editor>?

In an interleaved 60i video (normal old fashioned video), the image is scanned 60 times per second, producing one half frame at each scan (say the odd scan lines, followed by the even scan lines in the next half frame). Because movement can occur between the two half frames, you some times see interlaced jagged edges.

A better way is to take 30 still pictures per second of the entire image.

When your camera records at 30p video, it takes a single image – but splits it into two half frames and stuffs those into two 60i half frames (without any jaggies since its splitting one image in to two pieces whereas 60i creates one image from two separate pieces taken 1/60th of a second apart).

Consequently, a 30p video is stored as a 60i video. And two consecutive half frames, put together, become 30p. Your video software can’t tell the difference between between 60i and 30p.

So why do we have these strange 60i half frames? Historic reasons. The earliest TVs were not able to scan the full image top to bottom before the next image would arrive. The solution then was to draw only half the lines in each interval. Thanks to the persistence of the phosphor image of old TVs, the first lines remained glowing while the TV then scanned the alternate lines.

No one would design a TV like that today – but we’ve lived with it for many decades and it is still supported for compatibility reasons.

I can import 24p video into iMovie (or other editor) but it plays weird – why?

This depends on the camera. But it is common to store the 24p image in half frames, similar to 30p. But since 24 does not evenly divide into 60, the sequence of half frames is a little different.

In 30p, the sequence is basically 2:2:2:2:2:2 and so on where the 2 signifies 2 half frames.

In 24p, the video may be stored in various combinations such as:

2:3:3:2 or 2:3:2:3

In this way, the 24p mode uses up more half frames – consider 2 half frames and then 3 half frames. This slows down the video frames to match the 60i storage of the tape. Consequently, 24p gets mapped into a funky sequence of half frames on the 60i tape.

To a program like iMovie, this 24p footage looks just like 60i. But unfortunately, iMovie (And many other video editors) have no way of knowing that it is not really 60i footage with some frames appearing 3 times. So play back and edits produce strange artifacts.

If you want to convert this to proper 24p footage and then edit in an editor that supports 24 frames per second see these instructions.

As you may know, YouTube now supports both HQ (high quality) and HD (high definition) videos.

To create an HQ video, you need to upload a video file encoded at a rate of at least 1.5 Mbps.

To create an HD video, you need to upload an HD video file encoded at a rate of at least 2.0 Mbps.

YouTube recommends transcoding your HD video to 1280×720 prior to upload. This does work and produces a good quality HD conversion.

But here’s a little secret that I verified with some tests – don’t use 1280×720. Instead, transcode your video to 1920 x 1080. This produces the best YouTube HD video quality if you encode with sufficient bits.

Here is a video I shot in HDV (1440 x 1080/24p) – and after editing, transcoded to 1920 x 1080 / 24p H.264 at 5 mbps. The resulting huge file was then uploaded to YouTube. Watch it in “HD” and “Fullscreen”. Looks pretty sharp, heh?

(Sorry about the long zoom camera shake at the beginning – I was also dealing with someone else’s medical emergency as I tried to keep my mind focused on shooting this.)

Another trick for online use, if your camera supports it, is to shoot at 24p – that is 24 frames per second, progressive mode. There are three advantages to do this:

1. You avoid the interlace jaggies by shooting in progressive mode
2. You have fewer frames/second to process on your computer when re-encoding in to h.264 for upload, so this step is 20% faster
3. There appears to be a slight improvement in image quality.  While h.264 does interframe compression across a group of frames, there is a slight advantage in having fewer frames for a given data rate – so your encoding quality is a little bit better.