Working from audio recordings

ometimes it happens; not too often, but it does. A translator is asked to work from an audio recording. The client assumes that a translator frequently does this; so doing it in any of the translator's working languages should be a piece of cake.

In view of the concerns expressed by first-timers about such an endeavor in translator newsgroups on the web, I take this chance to voice out a few ideas.

1. The job

The request might be for transcribing, translating, or both, from an audio recording.

It can never be overemphasized that one of the translator's duties is to educate their clients. Some clients assume that translators who do dubbing/subtitling/captioning work from printed text. Though frequent flaws seen on TV might support such reasoning, it is definitely wrong.

So, if the client asks for both, hold your horses! Ask what they need the transcript for. If it's just to have the translation, sell them out of the transcript! It is not necessary. You can translate directly from the audio.

A transcript is useful for a written record of what was said in the recording, for developing printed material from its contents, or for rebuilding a script for an eventual re-enactment, but definitely not for dubbing/subtitling.

Therefore, it is important to know what the client intends to do with the transcript or translation. If it's video for dubbing or subtitling, this kind of work requires special skills; make sure you have them before accepting.

2. The media

A recording might be on various kinds of media, and some of them present specific difficulties.

If it is on waxed cylinders, long-play records, 16 mm (or any other size) film, Betacam (or any other strictly professional) video tape, you don't have the least obligation to have the equipment to play it. Ask the client to have it transferred to suitable media. Open-reel quarter-inch audio tapes may belong to this class or not, as once in the past home equipment for them was pretty common.

Let's analyze the kind of media you might get to transcribe or translate from.

1. Standard audio cassette tapes

   This is a quite likely format, as anyone can inconspicuously have such a hand-held or desktop recording apparatus. Likewise, a bulkier unit might have been plugged into the local audio system during the recording, or used for transfer from other media. The problem is in playing it for your work. As we'll see later, these jobs requires a lot of play, stop, and rewind a bit; or at least a lot of play-and-pause. Standard cassette players usually have a small delay in executing some, most, or all of these operations, so you might miss a couple of words every time you use them on a recording that calls for a lot of short rewinds, which can render the job extremely tedious and tiresome. Likewise, it's too easy to rewind more than necessary, which makes the process more time-consuming than it has to be.

   The alternative is to use a transcriber, which is a cassette player (and often a recorder, too) controlled by foot switches, featuring immediate action on play, stop, rewind, etc. so that nothing is missed. However such machines are expensive, I saw some for US$ 300 in the USA, and one should consider whether they would be used frequently enough to justify the investment.

2. Micro-cassette audio tapes

   These offer exactly the same advantages, disadvantages, problems, and solutions (including transcribers) as the standard cassettes above. The only likely additional disadvantage is that the units are usually palm-sized, and their control keys therefore are so small that their frequent operation is even more difficult.

3. VHS video cassette tapes

   These are intended for both audio and video. The acronym VHS stands for Home Video System. Of course there are sophisticated units in this format, but these are quite expensive, and not usually found in every household. Unless you happen to have a VHS unit suitable for accurate video editing, every pause or stop will cause you to miss a few seconds; this will require a short rewind, however longer than the audio cassettes described before. My advice in this case is to transfer the recording to a more suitable format.
   

4. Audio CD

   This is a very convenient format, as most players, even the cheap ones, will feature immediate-action play and pause functions, as well as fast fwd-rev cueing. If you are lucky to get a client offering the audio files in this format, take them after having checked that your off-computer player actually has those features and they work well.

   I don't recommend using the computer CD-ROM drive to play them while transcribing, since you will have to mouse-click on the screen to start and stop all the time.

   One useful bit of information is that CD audio files (*.cda) are just slightly modified computer audio (*.wav) files. While it's generally not worthwhile to burn audio CDs from, say, cassette tapes, if you don't have an off-the-PC CD player, a solution is to convert the CD audio files into *.wav files on your hard disk. There is plenty of software, even shareware that can do the job. I use an old version of DART CD-Recorder (US$ 29.95 from http://www.dartpro.com) just because I have it and because it does a lot more than this; I bought it for other purposes. But the freeware Express Scribe (see below) does it as well.
   

5. Audio computer files

   These will usually be either *.wav or *.mp3.

   If you receive them via Internet, beware of *.wav files: they are immense compared to the playing time. To provide some reference, one CD-ROM can hold up to 80 minutes of audio or 700 MB, i.e. 8.75 MB per minute of playing time. Of course, there are ways to make a *.wav file smaller (mono, lower sampling rate), but not as small as MP3 with compression.

   So, if you are to download or receive by e-mail audio files, ask for them in MP3 format.

3. The recording

Considering the act of recording in itself, never take anything for granted. If the client says it's a professional recording job, it only means that someone was actually paid to do it, and/or the equipment used looked impressive, nothing further.

You might get a recording where the volume was set too low or too high. If it was set too low, of course you can amplify the sound, but when you increase the volume, the background (or any other) noise becomes equally louder. If it was set too high, you get distortion, and some words may sound mumbled.

Even if it is a professionally made movie, there is the risk of music and sound FX covering the words or phrases you'll need to hear.

Another problem is the speaker. A weird accent is enough to create a need for listening to each phrase more than once, sometimes to the extent of requiring some creativity for it to make sense.

Speed is also an issue. Even a good, clear speaker, if talking too fast, might make the job more difficult than it should be. Ifthewordssoundgluedtoeachother, this means that the speaker's mouth is faster than your hand, and you might need a mini-rewind at each and every stop.

Almost each of these problems can be solved, though separately. Digital audio makes it easier without requiring hi-tech audio equipment. To solve them, I use a program named Acoustica (US$ 29.00 from http://www.aconas.de/ - 30-day free trial), whose 3.1 version allegedly can also replace the aforesaid DART to convert *.cda files into *.wav and *.mp3.

Let's assume that the recording you are expected to transcribe or translate is already in *.wav or *.mp3 format. (If not, you can convert it with Acoustica as well.)

The program will show you the recording in graphic format, so you can adjust the volume up or down as needed. You can even do different things to specific parts of the recording. One such case is when the person holding the microphone is close to the speaker, but just turns it around to get questions or comments from members of an audience farther away.

You can also remove noise rather easily. Noise may be a hiss from the tape itself, a hum from poorly grounded equipment, or just wind from a fan or air conditioner hitting the microphone directly. Just select a part of the recording (the longer the better) that should be silent, ask Acoustica to perform a "Noise analysis", then select the whole recording, and ask it to do a "Noise reduction" from that analysis. This should cover 90% of the cases. If you botch it up, there is always the "Undo" feature, and the chance of doing a somewhat "lighter" noise reduction.

Nevertheless, take care and listen to the "silence" you selected. Once I selected a piece of (what I though was) silence, but it included the speaker clearing his throat. Everything he said vanished in this operation!

Acoustica can also perform a "Time correction", increasing the total time of the recording without a corresponding change in pitch. So you can get the speaker to talk slower, which should make it easier to understand or, at least, give you an additional split second to stop playing and write it; when you restart, it won't begin in the middle of the first word. (Express Scribe does it too--see below.)

Beware of removing music and sound effects with this feature; there is a high risk of removing parts of the speech altogether. Better try to live with this kind of noise.

In time, I am not endorsing Acoustica above any other software you may have; it's just the best solution I have found for myself so far. Your mileage may vary.

4. The process

Your working method will depend on your personal memory buffer, not the one in your computer. Some people (probably all who do simultaneous translation) will be able to keep long phrases in mind for short recall. Others (like me) will only store short pieces.

This will determine how long you can listen before you have to stop and type. I know people who can listen to 20-30 seconds of a recording, stop, and then go hammering the keyboard without missing a thing. I stop at every phrase or, if it is long, at every punctuation mark. Test yourself, and find the best way for you to work. It's intuitive, but remember that you have to find your way, and not to learn anything new, though you might improve with practice.

A very useful tool for transcribing *.wav, *.mp3 and many other audio files (audio CD tracks included) is a program named Express Scribe (freeware from http://www.nch.com.au/scribe/index.html). You might even buy or build your own foot control for it; otherwise you can use programmable keys on your keyboard to play, stop, and rewind. It also offers variable playing speed for slowing down fast talkers.

I had been using electrically-controlled (immediate start/stop) quarter-inch tape open-reel recorders for this kind of work (actually translating videos for dubbing) for 15 years before I got Express Scribe. Quite frankly, I felt no difference between them and using Express Scribe; the switch-over was immediate and effortless. However such bulky open reel recorders have been discontinued for ages.

5. The price

Finally, the big issue... how much should you charge for such a job? It is not so easy to find market standards, though they exist.

There are two, totally incompatible, measurements for this kind of job. Each one burdens a different side of the deal with risk.

One is to charge per recording time (per minute, block of 10 minutes, or per hour). The risk is on your side. Before having listened to the tape, it's impossible to say how fast people speak on it, hence how long the text per minute will be.

The other is to charge per word. The risk will be on the client's side, as they won't know the size of their bill until the job is finished.

Whatever you and the client agree to should prevail, as long as both parties know what they are getting into.

One way of finding out the basic amount you should charge is by testing. Get an "average" or "typical" 10-min recording and do it! Use a stopwatch to time how long it took you to do the job, and count the words. Knowing how much you think you should make per hour, you can easily calculate the rates you should charge per word and per minute of playing time.

But this is not all of it. You have to check if it will be a transcription or a translation. Consider the additional time in research or whatever it will take you to do the latter, and add it on, possibly as a fixed percentage. Round this percentage up if you feel that there is a probability in spending extra time to understand a strange accent.

Also take into account whatever audio witchcraft you might have to perform, and add an adequate amount to cover it as a risk, not as an ever-present cost. Don't forget to include a marginal return on your investment in software and hardware, if needed.

Then perform a reality check on your final price, and adjust it. There is, of course, some risk that you will lose money (or get grossly overpaid and lose a client) in your first such job. But practice makes perfect, and while improving your skill you'll quickly discover what a fair price, which is both competitive and profitable, would be.