An example of using SGML for audio description

First, I assume you have read the introductory material on SGML for access, and that your curiosity is so intense that you need something approaching a real-world example.

OK. Here goes.

Let's think of the Descriptive Video Service audio-described version of Schindler's List. I'm going to give you a couple of versions of a passage from that film. (These transcripts were originally created for another project examining controversies in audio description, so you may consider them risqué. If someone would like to transcribe and annotate some other snippet from another DVS Home Video title, let me know. For now, that extra work isn't worth it for me.)

Ordinary text version begins:

Woman: Uh... all right.

Schindler: Please. [Motions that she sit down]

Narrator: As painters refurbish his factory office, Schindler has a young woman sit behind a typewriter. Schindler folds his arms and stares at her smooth, pretty face. [Types away] Another girl, a brunette, sits at the typewriter and smiles. Now, a tall blonde. Schindler kindly slides back the return carriage for her. She grins. Next, a young lady squints and leans close to the paper. Another smiling blonde. A thin, wavy-haired brunette with a coy smile. Schindler leans on the desk and gazes at a stunning young woman, who returns his unwavering stare. Now, a despondently-slouching Schindler as a gruff middle-aged woman types away, a cigarette dangling from her lips.

Stern: You need a secretary. Pick one.

Schindler: I don't know how. They're all so... qualified.

Stern: You have to choose.

Narrator: Outside, Schindler poses with 20 beautiful secretaries.

Photographer: Big smile, big smile!

Narrator: A photographer takes a picture. [Scene changes]

Now, another scene:

[Pfefferberg knocks on doorglass]

Pfefferberg: Herr Direktor?

Schindler, quietly, to Ingrid: Shit. I don't believe it. [Raising his voice] Stern, is that you?

Pfefferberg: No, it's Poldek. It's about Stern.

End of example. You note here that even embodying and audio-described film in print requires information that you don't find in the original, like [Pfefferberg knocks on doorglass].

Now let's consider how we would mark up this text for structure and not overt form. We find these kinds of structures in this example:

• Original dialogue
• What the DVS narrator says
• Notation of sound effects and action specifically for this written medium as I mentioned above
• Emphasis
• Pauses (notated by an ellipsis, ...)
• Manner of speaking: Quietly, to someone, etc.
• Exact pacing, which in a real example we would indicate by precise timecodes

So we could use markups like this:

• For dialogue, <dialogue NAME> </dialogue NAME> where NAME is the character's name: <dialogue Schindler> <dialogue Pfefferberg>
• For descriptions, something like <description> </description>
• For sound effects, <sound effect> </sound effect>; for action, <action> </action>
• For emphasis, the HTML tag <em> </em>
• For pauses, <pause> (doesn't need an off tag)
• For manner, <manner SPECIFICS> where we fill the SPECIFICS in, like <manner shouting> <manner to himself> <manner mumbling>

So the text example, when fully marked up, might look like this:

<dialogue Woman>Uh<pause> all right.</dialogue Woman>

<dialogue Schindler>Please.</dialogue Schindler> <action>Motions that she sit down</action>

<description>As painters refurbish his factory office, Schindler has a young woman sit behind a typewriter. Schindler folds his arms and stares at her smooth, pretty face. <action>Types away</action> Another girl, a brunette, sits at the typewriter and smiles. Now, a tall blonde. Schindler kindly slides back the return carriage for her. She grins. Next, a young lady squints and leans close to the paper. Another smiling blonde. A thin, wavy-haired brunette with a coy smile. Schindler leans on the desk and gazes at a stunning young woman, who returns his unwavering stare. Now, a despondently-slouching Schindler as a gruff middle-aged woman types away, a cigarette dangling from her lips.</description>

<dialogue Stern>You need a secretary. Pick one.</dialogue Stern>

<dialogue Schindler>I don't know how. They're all so<pause> <em>qualified</em>.</dialogue Schindler>

<dialogue Stern>You have to choose.</dialogue Stern>

<description>Outside, Schindler poses with 20 beautiful secretaries.</description>

<dialogue Photographer>Big smile, big smile!</dialogue Photographer>

<description>A photographer takes a picture.</description>

Now, another scene:

<description>Later, in his bed, Schindler kisses Ingrid, the flirtatious girl from the club. They roll around naked and entangled in the sheet. Out in the hall, Pfefferberg wanders toward the bedroom. Ingrid rolls to the top, her firm young breasts above his broad chest. Pfefferberg glances in and takes a bolstering breath.</description>

<sound effect>Pfefferberg knocks on doorglass</sound effect>

<dialogue Pfefferberg>Herr Direktor?</dialogue Pfefferberg>

<dialogue Schindler><manner quietly, to Ingrid>Shit. I don't believe it. <manner loudly, to the door>Stern, is that you?</dialogue Schindler>

<dialogue Pfefferberg>No, it's Poldek. It's about Stern.</dialogue Pfefferberg>

End of example.

Notice that all sorts of typographic niceties, like colons and spaces after people's names, are left out. In SGML, the program actually interpreting the SGML markup provides these according to whatever scheme you want. Also, some tags are clearly for direct human interpretation, like <manner quietly, to Ingrid>, while others can be automated, like turning <em> into italics.

We could take this marked-up file and produce all kinds of things-- a nice WordPerfect file with bold, italics, fonts, special characters (such as ..., ellipsis), and the like, for example, or, if we had timecodes from the actual source videotape, we could produce a script for a narrator and a description team or an edit decision list for a computerized editing terminal.

Clearer?

Back to the Joe Clark main page, or to the SGML for access page.