Notes localizer. Voice acting in games

As promised in previous articles, I continue the series of localization notes. This time it will be about sounding games.

At a time when localization was just taking its first steps, one could only dream of recording studios with professional equipment, sound producers and other amenities. And nevertheless, quite decent results were born that pleased the hearing not only with the original translation, but also with decent sound quality.

When we received the first orders with the requirement not only to translate the game, but also to voice it, we had to take urgent measures to resolve this issue. In the office room where we worked (a room of about 40 sq. M) a saw rustled, a drill was buzzing and a couple of drywall sheets tightly hidden one corner from our eyes and extraneous noises that could spoil the whole raspberry. At least it seemed so. Nevertheless, the noises were constantly trying to get inside, so a sign with huge letters “There is a voice acting!” Was placed on the door, which called for the lowest possible noise emissions. So the kettle and the microwave waited for a break, and even the sound of the keys at the computer tried to be quieter. And on the occasional “Aaaa-p-chkhii!” Instead of “be healthy”, something like “... your mother ...” was heard from the booth.

But not only the noises made the sound work difficult and tedious. Corner, with a voice acting booth, huddled on the heel size of not more than two square meters. Therefore, it could accommodate only a small table, a monitor, a keyboard, and an entertainer. Then, in order to speed up the work, I also coined it there, and let's call it, the “sound producer” who helped to select the necessary files for voice acting. So it was cramped, hot, and dark in the booth (the monitor and the desk lamp served as lighting). When the doors were opened about once every half an hour and people came out into the light, there was a feeling that they had spent at least a working shift in the mine face — soaked, squinting at the light and deeply breathing with life-giving air.
')
Before turning to more technical details, I would like to especially note the artists who so heroically endured the difficulties of work. While budgets allowed, we invited actual theater artists who really knew how to control their voices and how to give the characters a special flavor and character. Unfortunately, the financing of projects was not large enough to voice all the roles by different people, so usually each artist took on several roles. But how did they do it! I still can not forget the talent of Cyril, who knew how to change his voice that freely parodied Doctor Livesey from the cartoon Treasure Island, voiced several characters so that it was impossible to determine what the same person said, spoke Chinese so that the Chinese They took him for their own, although he simply gave out a memorized set of sounds.

In general, I take off my hat. This is not easy. A few years later, when I had the opportunity to launch another “studio”, with more modest budgets, I had the opportunity to test for the role of the sound students of the theater institute. A whole group of 20 people responded amicably to the announcement of the work :) But in the process it turned out that not everybody could read a paragraph of text in front of the microphone with expression. People strayed, lost their breath, or there was just an inexplicable fear of the microphone - the person who had a great role in the theater simply could not speak knowing that they were recording it. Mystic.

It's time to pay attention to the technical side of the issue of sound.

For a start, of course, it is important to place where the voice is. As I said, drywall is not very helpful in soundproofing. The next "studio" was covered with a thick layer of paralon. Inside it was quiet and ... soft. Thoughts of an insane asylum involuntarily appeared, especially considering that the door was locked from the outside. In such a room, everything is perceived differently. There was a funny case - the game tester checked in this room for the passage of Silent Hill 4, and four times in a row from and to ... After that, he must have been afraid of sudden movements and sounds for about a week.

Equipment. Variations are possible here. In the beginning, it was a regular computer with a Sound Blaster sound card. She has a low noise level when recording. Later, there was a special mixer, which took on the task of setting sound levels, adjusting sensitivity, and so on. Particular attention should be paid to the microphone. If in the 90s a regular microphone from a tape recorder could also be used with proper skill, then later the purchase of a professional microphone for voice recording is simply a necessity. Moreover, as it turned out that the variety version of the microphone behaves much worse than the special announcer with wind protection. Forgive me sound industry experts for such an amateurish presentation, I personally am not a super expert in this area and I don’t know how to discuss the frequency characteristics of microphones, amplifiers and other shamanic attributes for hours :) So I’ll just say that if you want sound to be good, consult a professional about the technology. And better, of course, go to a normal studio and do not worry about your brain.

Certain difficulties arise when reciting the translated text. Usually in games, the sound is broken down by file. One file - one phrase. And well, if the game itself determines the duration of the phrase by the size of the file time. It is worse if the playing time is rigidly tied to events in the game, or it is simply written where it is not known. Then the sound engineer has an interesting task to read the text so that the Russian version fits in during English. Which is often quite difficult, since the Russian phrases themselves are a bit more English. If the translators “forgot” about this moment and translated everything “in a big way”, then everything goes in the second circle. Texts go on correction and shortened. It is necessary to have a flair to read a paragraph of text to calculate the speed of speech so as not to go beyond the limits of a given time. But do not be in a hurry - otherwise the character in the game will still be destroying, although the phrase has already been said. Given these nuances, the total time of sounding is not even close to the material time. Speaking of unfortunate five minutes can take half an hour of time. Also, small files are pulling in the process, purely physically. The file must be opened in the program, find a translation, read out, save. It would seem that the simplest operations, but on a large number it is just awful. I remember the voice of hockey, where there were about 20 thousand of the same type of short phrases of the commentator. Got probably a week. It was just torture for the artist, when from the booth came the “First, first, first.” The second, second, second ... "and so all the numbers of the team, and then the declination of numbers and players and other combinations of phrases. I do not know who came up with the developers, but the sound engineer went on a smoke break with a cup of coffee and continued to mumble phrases on autopilot. Although it may just quietly swear ...

In order to reduce the time for opening and saving files at least a little bit, we then applied the technology of gluing all small sounds into one “sausage” with markers for Sound Forge. As it turned out, it was more convenient for dubbing and translation. Conversely, by markers, all this was programmatically cut and inserted into place.

Sometimes there were interesting problems with translation. For example, when a phrase has one beginning and several different terminations. Although it would seem, why did developers save on sound so much? I had to invent a Russian version of the phrase. What is not always easy, given that the phrase in the game could refer either to a male or to a female character, depending on the game situation.

Another problem, when you need to voice the video by the "vojsover". When the whole game is in Russian, and in the video on the background, English is heard, it is somehow not very good. Often it turned out to make the so-called "backing track". A feature of the studio recording of a stereo track is something like the inversion of sound in the channels (again, let the specialists forgive me). The voice is written as though in the middle. And if you make an inverse overlay of the left and right channels, you get a mono sound free from the speaker's speech. Hmm ... the explanation doesn't claim to be an Oscar, but the technique is simple and done in three movements in Sound Forge. Who will be interested, rummage in the search engine.

There are still quite "banal" problems. Strangely enough, game developers did not always use standard sound formats like ogg, mp3, wave in their creations. There were always clever men who invented their own sound format and their own compression. Therefore, over the years of work in this area, we managed to learn a lot of interesting things about digital audio coding, variations of ADPCM compressions, and write our own codecs for sound packaging. As it turned out, it is not always easy. It seems that you make your converter from the usual wave to the game format, and in the game some clicks appear in some places of the phrases. It turns out you need to add filtering form, make sure that the sound level does not fall out of certain limits. After all, if a normal sound uses 16 bits of information, then in ADPCM, it is usual to squeeze it into 4 bits depending on the variety. Now for those who are interested in this area, the sox library is available in which you can learn a lot of useful algorithms.

Perhaps this time is enough.

Wait for the next issues.