Video Stabilization

N93 Video Stabilization:

After experimenting with the video stabilization feature built in to the Nokia N93, I don’t think it is a good idea to use it because of the jerkiness it adds to pans. And it doesn’t help nearly stable shots as far as I could tell.

After Effects Video Stabilization:

If video needs to be stabilized, it is better to do it while editing. I got reasonable results with After Effects. In the first clip, the camera was mounted, but there is shake when the door slams. In the the second clip, I reduced the wobble in a nearly stable handheld shot.

To stabilize with After Effects, import the clip and drop it into the timeline. Select “Stabilize Motion” in the “Animation” menu. Make sure both “Position” and “Rotation” but not “Scale” are selected in the tracker controls pane. Place the current time indicator at the start of the section needing stabilization, and place the track points on small objects that don’t move. Click the play button next to “Analyze:” in the tracker controls pane, and click the stop button when the indicator in the composition window gets to the end of the section. Then click “Apply” in the tracker controls pane. Next select “Transform | Scale…” in the “Layer” menu, and increase the scale just enough to cover any black regions that appear on the edges of the video. If the results look good export.

Sound for Everio Video Clips on the Mac

Since I switched to editing video on the Mac, I wasn’t getting any sound with clips I shot with my JVC Everio camera. I assumed I made a mistake during recording. The battery in the shotgun mike can be dead, or I can forget to turn the mike on. I noticed there is a green indicator of sound level on the view screen. Also, the camera does have a tiny speaker, so there should be faint sound during playback.

On both Windows and Mac, it is necessary to rename the .mod files to .mpeg in order to play them. On Mac, they won’t have any sound when played by QuickTime or edited by Premiere. You can get the sound by running the .mpeg files thru a downloadable utility called DVDrop, which converts them to a .dv format. The video quality is degraded, so I just use the .dv files for audio.

The best workflow is to assemble the video in Premiere using the .mpeg clips. Then import the .dv clips into Premiere and run them thru Soundbooth if necessary. For the video above, I used the default noise remover, which made a large improvement, and the default clicks and pops remover, which didn’t do much. It is important to re-render the sound in Soundbooth before trimming the sound clips down to size, because otherwise audio transitions to smooth out the sound at cuts won’t work.

The shotgun mike might not be ideal at close range. Little things like a shoe scraping on rock sound unreal. I was using the middle setting on the mike; maybe the higher setting would dampen the scraping noise.

Update: 28 December

I changed my mind about the best workflow. Drag a new .mpeg file and the corresponding .dv file into the timeline. Remove the video of the .dv file, and then position the remaining audio under the .mpeg file and link them. Since neither .mpeg nor .dv file has been resized, the synchronization should be perfect.

Premiere Pro Exporting and Importing Notes

When exporting from Adobe Premiere Pro to YouTube use

File | Export | Adobe Media Encoder...

Set the format to QuickTime. Remember to check the Export Audio box, which is not the default. Set quality to 100, frame rate to 30 fps, and frame dimensions to 640×480. Set the bit rate to 6 Mbs. You may need to use a lower bit rate if the video is longer than 2 minutes to keep the file size under 100 M.

YouTube recommends 320×240 frame dimensions. Odd because the YouTube player looks like about 450×337.

I had difficulty importing .mod files created by a JVC Everio camera into Premiere Pro. The .mod files can be imported into Premiere Pro or played by QuickTime if the suffix of the file name is changed to .mpg. However, Premiere Pro inserted some glitchy red frames or red splotches, forcing me to edit some of the footage with iMovie.

New Software

I edited my first video with Premiere Pro, using hand held footage from my phone. The clips from my phone have an .mp4 suffix. This format imports into Elements, but not Pro.

The Wikipedia article says that .mp4 is “essentially identical” to .mov, but just changing the suffix isn’t a solution. QuickTime plays .mp4, but not if the suffix is changed. I upgraded to QuickTime Pro and used it to convert the video. I didn’t notice any degradation in quality.

I installed ffmpeg, which can be done with MacPorts. .mp4 and .mov are both supported formats, but I haven’t figured out what the command line arguments need to be to perform the conversion.

I exported from Pro using the default format, and it created a 200M file. Too big to upload. Whatever format that was didn’t allow the quality to be adjusted. So I exported in .mov format. I got a 55M file and uploaded that.

I made my own soundtrack with Garage Band. Now I’m thinking I need a better microphone, a midi keyboard, and more samples. I don’t want to sink a lot of time into music making, though.

To do: figure out what the ideal export settings are. The video doesn’t have 4:3 aspect ratio, and 10M-20M per minute is best for uploading to YouTube.

Removing Objects from Video

Twice I’ve been able to remove people who wandered into the background of a shot. This technique only works for stable shots taken with a tripod. The instructions assume editing with Adobe Premiere Elements.

Take the portion of the clip without the unwanted element and drop it into the layer above the video with the unwanted element. Time stretch the top clip if you need more length. Then take a crop video effect and apply it to the top clip. By setting the values in the property window, crop the top clip down to a patch which covers only the unwanted element.

A challenging situation is where the person in the foreground passes in front of the unwanted element. You can split the top layer into two or more clips by placing the current time indicator and clicking the split clip icon in the monitor window. Alter the cropping clip by clip so that it never covers the person in the foreground. Use a cross dissolve on the top clip to reduce the visual impact of bits of the unwanted element you missed.

Additive Dissolve Demonstration

Maybe an answer for Chris V:

Additive dissolves don’t guarantee that there will be a totally white frame in between the shots. The new shot fades in completely, before the old shot fades out. At the midpoint you will have white where you have white in either shot, or where the colors from the two shots are complementary and add up to white.

In a cross dissolve, the old shot fades out as the new shot fades in. The amount of brightness will never be greater than the brighter of the two shots, so you don’t get the flash of light that you get in the additive dissolve.

I think additive dissolves work well with shots that contain a gradient, which you will always find in blue sky.

Transitions

The natural eye can’t cut from one shot to the next. Hence long shots are better at creating the illusion of being there. On the other hand, long shots usually drag, so an editor is forced to make cuts. There is never any reason to draw attention to cuts with goofy transitions, though.

The best cut is a straight cut, but it doesn’t always work. Cutting from an unstable shot to a stable shot looks bad. Even for stable shots, the cut will be jerky when similar objects are in approximately the same location. The problem can be corrected with a cross fade or cross dissolve. Compressed video with cross fades will pixelate, especially on Youtube, so it is good to make the cross fade fast. The default in Adobe Premiere Elements is 1 second, but experiment with half a second.

I like the additive dissolve. It’s like a dip to white, but the flash of light is irregular and energetic. I picked it up from this music video, where it works well with sunlight and blue sky.

I like the wipe. I had troubles using it with the first version of Premiere Elements. Elements would freeze one of the shots part way through the wipe. Actually, I’ve had this problem with dissolves as well. It’s possible that both shots need to be trimmed with usable video in the trimmed off area. The problem is not as severe with Elements 3.0.

I’ve been watching this video over and over. The camera work is well beyond what I can do. Lots of cuts from the shot to the expression of someone watching the shot. Or sudden zooms on someone’s face, which mimic eyes dilating from adrenaline.

Nokia N93

The manufacturer spec states that the N93 doesn’t support miniSD cards larger than 2G, which explains why the 4G card didn’t work. I have a 2G on order from Amazon.com. This should be plenty of space. I shot a lot of video during my holiday trip, but didn’t even fill up half of my 1G card.

I lost my lens cap. I can’t find them for sale, so I will cover the glass over the lens with clear adhesive plastic.

Will I ever be able to upload movies from this phone directly to the internet without using a PC? Most carriers in the US are using so-called second generation technologies such as GSM, a digital standard which guarantees 14.4kbit/s of error corrected bandwidth, too slow to transmit video.

I need a service provider using a third generation standard such as UMTS (also called WCDMA), which typically provides a bandwidth of 384kbit/s. Cingular is currently the only carrier in the US with UMTS. However, Cingular does not support the N93 because it does not support the 850 band. The N93 can operate via GSM in the 900/1800/1900 bands and WCDMA in the 2100 band.

T-Mobile, which is my carrier, recently won an auction for bandwidth in the 1700/2100 bands, and will be deploying UMTS in those bands in 2007, so I look forward to being able to upload soon.

Tonight I will try to get my phone to communicate with my laptop via Bluetooth.

The Nokia N93

Last Monday I picked up my new Nokia N93 from UPS.

For the most part I am happy with it. It is bigger than my previous phone, about the size of most cell phones from the late 90s. So I’m wearing it on my belt, instead of in my pocket.

The 128M minSD card that comes with the phone is probably insufficient. It uses about 30M for every minute of video shot, and lots of extra memory is needed to edit video.

I purchased a 1G card from Best Buy for $50. I also ordered a 4G card online for $100. However, the phone isn’t recognizing the 4G card. Is it a bad card, or is there an incompatibility? SanDisk is not the manufacturer of the 4G card.

The video is good. It may approach DVD quality. I need to burn some video to disc to test this. Of course there is a loss in quality when the clip is uploaded to YouTube.

Editing movies on the phone is doable. It is easy to arrange the clips in the order you want them, and to trim the clips down to size. Trimming is done by watching the clip play, and setting the start and end marks as it plays.

The editor allows you to create a title screen, and the choice of transitions include crossfades, wipes, dip to white, and dip to black. There doesn’t appear to be a way to adjust the size of the text in the title screen, and it is too small for YouTube.

I have yet to use the sound editing. You can mute the audio on your clips and insert an mp3 sound clip.

Editing can be frustrating. If you save the video and then re-open it in the editor, the editor will treat the video as a single clip, so there isn’t much motivation to save while editing. The danger here is that if you accidentally hit the “end call” key, you will immediately exit the editor with your work lost. Also, I had a difficult time figuring out how to enter the text for a title screen. After experimentation and losing several edit sessions, I discovered that hitting the left select key once after entering my text works.

Some of the software or perhaps the OS is still glitchy. On several occasions I found it necessary to cycle the power because the phone was hanging.

HDV

The weekend before last Harlan told me that he had the opportunity to buy a professional grade HDV camera for less than $1000. He asked me whether it was a good deal. It seemed like a steal. However, I didn’t know much about editing and distributing HDV; there could be additional expenses, depending upon what wanted to do with the video. I decided to research the topic.

formats

The first question is what format the camera uses. There are high definition formats other than HDV, such HDCAM, but HDV is the only format in the price range of consumers.

what is high definition

High definition is anything with more that 576 horizontal lines of resolution. Optionally it is anything of higher definition than the common analog TV formats such as NTSC and PAL. Color NTSC has 484 horizontal lines delivered at 29.97 frames per second. NTSC is an interlaced format, so half of the lines are refreshed every 59.94 seconds. To compute the number of pixels, use the fact that NTSC has a 4:3 aspect ratio.

The common HDV formats are 720p and 1080i. The number indicates the number of horizonal lines, and the letter is an abbreviation for progressive or interlaced. If the format is progressive, all the lines are updated each cycle. Frame rate can be set to 24 fps (film), 25 fps (PAL), or 30 fps ( approximately NTSC). Aspect ratio can be set to 4:3 (TV) or 16:9 (film). Data is compressed using MPEG-2.

cameras

HDV cameras for the consumer can be had for less than $1500.

a review of HDV cameras by JVC, Sony, and Canon

editing

As of version 3.0, you can import HDV into Adobe Premiere Elements ($100). The extra resolution is lost. This may seem purposeless, but it does allow you to use your HDV footage now, and re-edit your footage with a true HDV editor in the future.

Options for editing native HDV include Adobe Premiere Pro ($700) and Final Cut Pro (versions 5.0 an later, $1300 as part of Final Cut Studio) . The most cost effective solution appears to be Pinnacle Studio ($120).

computer

I am able to edit short videos in DV with 500M of memory. I doubt this would be sufficient to edit HDV, which will have 10x file sizes.

distribution

I don’t know of anyone distributing HDV over the internet. Youtube has about 1/30 the bitrate of a DVD. Of course Youtube screens are smaller, but even per pixel the bitrate is about 1/9 of a DVD.

Hence, to distrubute your HDV movie, you will be forced to burn discs. You can use Pinnacle Studo to burn in an HD-DVD format on regular DVDs using a DVD burner. These discs cannot be played on a regular DVD player, but they can be played on an HD-DVD player.

Optionally you can use Roxio DVDit Pro HD ($500) and a blu-ray burner ($750 and up). Both options are described in a MediaLive article.

playing your disc

You will need an HD-DVD player ($500) or a Blu-Ray player ($1000). You will need a high resolution TV (30″ screens run about $800).