Video Editing

Shooting for the Web

Special considerations for technology's limits

Almost everything about shooting video for the web, as opposed to shooting video for any other display medium, is about respecting the limits of current technology and learning to use those limits in the best way possible. Without this knowledge there is a very real Alice in Wonderland wishfulness about creating video for the web.

Video on the web is limited by the amount of data which can be delivered per time period - bandwidth. There are three key links in the data chain: 1) the server sending the data, 2) traffic on the internet across which the data is sent, and 3) the client's ability to receive quantities of data.

All too often we run into a David (the client) versus Goliath (the huge media file) contest simply because the creators of the media file don't get:
1 - that their file size is too large for anyone they want to reach to be able to see
2 - that the internet cannot support theater or TV quality no matter how hard you wish it might
3 - that the equipment and bandwidth specs are measured under ideal conditions and that you never really get the full-rated bandwidth
4 - that the internet is a shared highway subject to traffic jams
5 - that each client is doing several things on their one internet connection: your video/audio, their email, instant messenger, other web pages and Google, Yahoo, etc. as well as applications from word processing to spreadsheets and so forth.
*** It all has to fit into the same space.

The bandwidth you want your piece to have.	The bandwidth you actually have.
wishful thinking	reality

Do's and Don'ts of Web Video

REMEMBER: These are not intended as creative rules. All the rules in this section derive from the currently limited ability of web media to deliver amounts of data. Following these rules will let you place media on the web which is actually usable and to which someone will return more than once, we hope. Remember also, limits often force you to become creative within them. You will not be able to follow all of them, nor is that desirable. But they all provide better media delivery on the web.

These rules are similar to an early time in cinema when the limits of film and lighting produced similar techniques, though for different technical reasons. Studing old films is helpful to see how these shooting rules work.

How the GOP gets into every stream

GOP is an acronym for Grand Old Party in politics but not here. GOP for us means Group Of Pictures. One of the major ways that a media encoder tries to reduce the quantity of data being sent down the pike is to only send those parts of the video which change from one frame to another. First, the frame is divided into a series of small rectangles. You've seen these rectangles in compressed video by now. They make the frame look like a mosaic or give it a quilted quality.

Video is sent in segments. Each segment starts with all the information in the first frame (the index frame). For each succeeding frame only the parts of the frame which change are sent. A full frame is sent again, starting another segment, when one of two events happens, too much information in the frame changes at once or we have come to the intended end of the current segment.

This is why the less movement and the less detail in the picture, the smaller the data load when downloading. If you remember this reason the rules won't have to be memorized. You will be able to invent them on your own each time you need them.

The video (here illustrated using an animated GIF).	The first frame sends ALL the data.	Each successive frame sends only the items which change.

Reality Check - Actual Throughput

data in kilobits/second = (filesize_in_kilobytes * 8) / seconds_to_download

You really need to know what kind of experience your clients are getting. That is you need the receiving end of the experience. Don't depend on wishful thinking about what you hope they are getting from your site. If you are going to design a site don't do it based on what someone should be able to enjoy or what you should be able to show your visitors. In the words of an old saying: don't should on your visitors.

Use modem equipment - not broadband and not your local machine. Really download your own multimedia files. Be a typical visitor. If you watch the connection dialog during the download you will see that the actual download speeds varies enormously during the download. The final download speed, therefore is really an average speed. You will see that the claims for bandwisth speed never actually match reality. So design your multimedia delivery to meet real experiences.

Motion Control

Use a Tripod - NO HANDholding

Use anything to fix the camera in a stable, non-moving position. Mainly that means a tripod (good, stiff and maybe even heavy). The tripods with plastic tops often vibrate so badly you are better off hand-holding and you know what I just said about handholding. Use a bean bag if you have one. If not a pillow, rolled up coat or other cloth will help. Or even attach the camera to a table top or hold it against a wall.

You need to realize that the seemingly tiny movement you get, even with the very best handholding is interpreted by the streaming encoder as a new GOP (group of pictures) to send. Each seemingly tiny move means that the maximum data needs to be sent out again. A tripod eliminates those mostly un-felt moves.

Use SIMPLE Backgrounds

Plain backgrounds compress easier. Backgrounds with a lot of small detail require extra pixels to handle. Detail means more data. More data means more bandwidth is taken up.

Use Static Images (to the extent possible)

Okay, we are talking about video and that does mean that things move. So it really isn't possible to use only static images (forgetting about photographic illustrations). The real point here is to have as little in the video frame that moves as we can get away with.

Do NOT PAN OR ZOOM The Camera

This has exactly the same effect as handholding - in terms of data being sent. Each pan or each zoom changes every pixel in the picture meaning that each and every frame has to be transmitted as a new group of pictures (GOP).

Use Cuts Rather Than Transitions

Abut one clip directly up against the other. Make the picture change without any kind of blending between the frames. When the encoder sees the blended frames it means that every frame in the transition is totally different from the previous one. That means that each transition frame requires that everything in the frame is encoded for transmission. Everything. Transitions require far more bandwith because they send far more data.

What are some of the results of too much motion?

At the Encoder

The encoder may just "decide" to simplify the image according to its settings. The encoder may do any of these things:
1 - Often just "mooshes" the image into unfocused-looking blobs. Your nice motion looks as though it were shot through a heavily steamed-over window.
2 - May reduce the frame rate
3 - May turning the movie into a slide show.

So essentially you just lost control over your work because you didn't shoot or edit within the technical limitations of the output format.

At the Client

All of these are the result of too much data for the "pipe," which is too small.
1 - The client player, whether stand alone or embedded, may just stop.
2 - It may show a little of the file then wait for more to be downloaded (it keeps buffering)
3 - It may wait until the entire file can be downloaded and might not play until then (could be a lo-o-o-o-ng wait)

Getting the Exposure Right

 

Set the White Balance Properly

Setting the color balance for the lighting will go a long, long way to providing a clear picture. The wrong color balance (set the white balance) gives a muddy picture which loses detail.

Set the Correct Exposure - MANUALLY

You want a clear set of image details in the most important areas of the picture, normally, people's faces. So expose for the faces, as seen from your viewpoint, and forget the rest. If you have the fortune to have a camera in which you can set contrast levels, and you understand it, use it. It will allow you to create a long-scale across the picture. The scale is the range of tones from the blackest black to the brightest white in the scene. A long scale is able to differentiate more tonal levels - as opposed to black, a few grays and white.