Voice Acting Tutorial by dlw6

Voice Acting Tutorial
by Don Willadsen, known on TTLG as dlw6 and on Privateer Gemini Gold and VegaStrike as Turbo

Back to my Voice Acting PageSite MapLooking for web space?  Use what we use!

Special thanks to Gore_TornNameless_Voice, and Slyfoxx for suggesting improvements to this document.

Introduction & Tools

Voice acting adds depth to game mods by providing audio material which the author can use to create conversations and characters (referred to as characters when they appear in movie clips and as AIs when they appear in-game) that can provide clues, entertainment, or atmosphere for the player.  Good games have a good story, and voice acting can enhance the storytelling.  This tutorial will show you how to record your voice for use by game authors.nbsp;

You will need recording/editing software. 

  • Audacity is a free open source product that has useful features to make the process easier.
  • The Windows Sound Recorder is built into Windows 3.1, 95, 98, ME, 2K, and XP (maybe NT too, I’m not sure) and, while not as feature-rich as Audacity, still works.  If you have a Mac, it probably has something equivalent or better.

You will also need a microphone.  There are many types of microphone and many levels of price and performance.  You need not spend a lot of money.  You are recording your voice for use in-game and minor differences in quality will not be noticed by most players.  I started with a $10 microphone, the type designed for web teleconferencing, and it worked fine.   I later moved up to a $50 headset microphone and it worked slightly better.   You want a microphone you can put where you need it and have it stay there. The built-in microphones found in some monitors and laptop computers may require you to position yourself instead of the microphone, which can affect the quality of your voice.nbsp;

The last thing you will need is a script.  This is a description of what the character will say and how he/she will say it.  This might come from the mission author, you might write it, or it may be a team effort.

Setup:

You will have to experiment with different settings on your computer & microphone combination to find the best quality.  My old computer had Audio Equalizer software in which I have to turn off Noise Reduction on the microphone to avoid loud “pop” sounds that it causes when starting and ending recording, presumably from turning the mike on and off. My new computer lacks this, and so I have to remove such sounds manually (which I will describe later).   The room should be quiet, and the CPU fan and other sources of noise should be as far from the microphone as possible. 

You should experiment with the settings on the microphone and your computer’s audio card. If the volume is too high, your voice gets distorted, and if it is too low and your voice will be too quiet to hear.  The software can correct low volume more easily than distortions, so err on the side of lower volume, but if the volume is too low then the amplification process will cause distortion.  Some sound card software includes a “Mike Boost” option; if you have it try turning it on and off and see which sounds better.

Finally, experiment with placement of the microphone; for the desktop mike, 14″ from my face and a bit to one side worked best.  For the headset mike, I wear it with the mike near the side of my mouth.  Putting the microphone to one side helps eliminate the “pop” sound that can sometimes happen with hard consonants, such as “P” or “B” when your mouth emits more air. 

When you find a good setup, write the settings in a file and save it in your voice acting folder for future reference.

The recording software must also be configured.  The AI sounds in Thief 2 The Metal Age are 22 KHz, 4 bit, mono IMA ADPCM (a type of compression) format WAV files. The sounds for VegaStrike, on the other hand, are 44KHz 8 bit mono OGG.

To set this up in Audacity, use the menu Edit .. Preferences, and change the settings as follows:

  • Audio I/O tab, set Recording Channels to Mono.
  • Quality tab, set the desired settings; for example the Default Sample Rate to “22050 Hz” and default sample format to “16 bit” (the lowest it will go)
  • File Formats tab, set desired format, such as Uncompressed Export Format to “WAV (Microsoft 4 bit IMA ADPCM)”

This will set Audacity to export your recordings to WAV files at the desired quality. If you are voice acting for some other game, use the same procedure to set your software to match the original game’s sound file format.  To determine this qualify, right-click a sound file from the original game, select Properties on the pop-up menu, select the Summary tab, and note the format information you see there.

To set up the quality in the Sound Recorder, use the menu File .. Properties, click Convert Now, and set the Format to “IMA ADPCM” and the Attributes to “22.050 KHz 4 bit mono”  Use the Save As button to save these settings for quicker access.  At this point, keep Sound Recorder open so you do not have to reset the attributes again, and just keep saving new recordings under different file names.

When recording, I keep several windows open on the computer.  These include the script, Volume Control and other audio settings that your computer may use, and the Recording Software.  Resize all your windows so that your script and the recording software are visible at the same time.  You will be reading the script as you make the recording.

Finally, create a folder for the character or conversation you are recording.  This will help you organize your files as you create them, and publish them more easily.

Script:

The script is a description of what the character will say and how he/she will say it.  If you are doing a conversation for an author and a specific mission, the author may write it for you.  If you are doing a voice set, to be used in a more general way, you may be the writer.  The most enjoyable voice acting I have done have been those projects in which the author gave me generally what he wanted and I decided exactly how to act it out. 

For conversations, I always write out the script.  For voice schemas (the dozens of things the AI will say in various unscripted situations) I usually think about what the AI will say generally, then just record a lot of lines and keep what I like. 

You should rehearse the script until you are comfortable with it and can say it the way you want.  This is particularly true if there are unfamiliar names, difficult-to-pronounce words, or if you are changing your voice with an accent or other characteristic.

The Thief and SS2 games, for which I do most of my voice acting, use an engine in which conversations are set up as separate WAV files for each line, and the AI take turns speaking via a succession of WAV files scripted to play in a certain order.  This approach creates realistic sounds because the voice acting lines enamate from the AI who is saying them, so the player’s ability to hear the lines is affected by distance and terrain.  This tutorial is written with the “one line per file” approach in mind.

Recording:

Recording is, in technical terms at least, the easiest step.  Click the Record button on your software, say your lines, and click the Stop button.  My preferred method is to start recording, wait a second or two, and record every line for a single character with a pause of a second or two between each line, then wait a second or two at the end before I stop recording.  Do not feel rushed to get all the lines in, as you can remove all the silent parts during editing.  The silence helps you see where one line ends and another beings during editing, and lets you build a noise profile (more on that later).  Recording the entire conversation, or schema subset, for one single character at a time makes it easier to keep your voice consistent for that character.

Next, listen to the recording.  Listen for errors in what you said or how you said it. Listen for noise in the background.  Steady noise can be edited out, variable noise is harder to remove. Listen for popping or hissing from getting too close to the microphone, but do not worry about any such sounds that occur during moments of silence because they will be easy to remove later.  Ensure the volume of your voice is consistent and appropriate to the script.  If you like the result, save the file in the folder and proceed straight to editing.  If not, erase it and record again.

The method used to erase your recording, if you do not like the result, varies with your software.  For Audacity, you can erase the entire recording by clicking the “X” button in the upper-left of the audio track to be erased, or you can drag-select parts of the recording then press the Delete key.  For Sound Recorder, you can position your cursor at the desired spot by moving the slider or using the forward and reverse buttons, and erase either everything before or everything after that spot via the Edit menu.

Editing:

Editing is the most technically challenging step.  There are three things your recording may have that need to be edited out:

  • Noise:  background noise or other things you did not intend to be recorded
  • Blanks:  empty space at the beginning and end of each line is good to help you separate the lines, but silence at the front and back of the final WAV files makes the files larger
  • Recording artifacts:  some microphones may “pop” or “crackle” when you start speaking or when you start and end the recording, or may sound distorted when you say hard consonants like “B” or “P”

Both Audacity and Sound Recorder will show you the waveform that was recorded.  The main difference is that sound recorder shows you only the part of the waveform at a time, while Audacity lets you set a zoom level to see exactly the part you want to see:

There are two things the waveform can show you.  First, you will see in Audacity that there is some background noise, because when the actor stopped speaking the wave form did not go completely flat.  We will fix that later.  Second, pay attention to the amplitude (height) of the waves.  If the waves are full height, such that they touch the top or bottom of the space, that part of the sound is as loud as it can go and you should (1) consider turning down the volume on your microphone or repositioning it further from your mouth and (2) listen for distortion in that part so you will know if you need to re-record.  Ideally, you want the wave to approach the edges but not quite touch it.

If your recording sounds good and you do not see any problems, the next step is to remove the noise.  Sound recorder has no such feature, but Audacity has an easy process to remove noise.  First, select a portion of the recording during which you were not speaking (one of the silence/pauses you recorded between lines), then use the menu Edit .. Noise Removal:

Click the “Get Noise Profile” button listed under Step 1.  You only need to do this once for each file.  If you change your setup, volume settings, or microphone position, load a different sound file, or close the Audacity program and restart it, you will want to re-create the noise profile.

Then, select the entire recording with the menu Edit .. Select .. Select All (or CTRL-A), and click the menu Effect .. Noise Removal.  Move the slider to the desired position (I set it about 3/4 of the way to the “less” side, because removing noise also makes the recording quieter), and click the Remove Noise button.  You can also use the Preview button to hear a second or two of what it will sound like.  Here is part of the same sound shown above, with the noise removed:

Listen to the recording again to ensure you like it.  Save the file again, and you are ready to break the recording into individual lines.

With Sound Recorder, this can be tedious.  You have to listen, pause, move the slider, look at the wave form, and note the time to figure out where the sound starts.  If you use the mouse to select the slider, the left and right arrows will move it in 1/10 second increments, which helps you zoom in on the spot you want.  Once you get as close as you can to the beginning of the sound, use Edit .. Erase All Before Current Position.  Listen to the file again to ensure you did not remove anything you need to keep.  Then, find the end of the line you want using the same process and Erase All After Current Position.  Listen to the line again to ensure it sounds okay.  Save the file under a new name, reload the original recording, and repeat the process for each line in the recording set.

Audacity makes this process much easier.  First, drag the mouse to highlight the sound you want (the highlighted area will show as a darker gray per the screen shots below).

Then use the zoom button to enlarge that area, using the “Fit Selection in Window” button found next to the other zoom buttons.  After zooming in, you can easily use the mouse to drag-select the sound more exactly without any blanks at the front or back:

Use the Play button to listen to the highlighted area only, to ensure you selected exactly what you wanted.  If the noise reduction was insufficient and there is still noise where there should be silence, Audacity has one other feature that will help. Carefully drag-select a silent part, then press Control-L to render that portion silent. While this will not solve noise that occurs during the speaking parts, it is a good trick to remove noise during times that should be silent, such as when you inhaled between phrases or smacked your lips between sentences. Besides, background noise that occurs while you are speaking is much less noticeable than when you are not.

Finally, use the menu File .. Export Selection as WAV to save that part of the sound as a WAV file.  Because you already set up the defaults in your software, the WAV file will be saved in the desired quality. 

The voice files for any particular character should have names to indicate what character they are for.  Make sure the WAV file name you select follows a pattern similar to the game for which you are recording the voices, and that your file names are not the same as any file names from the original game.  For example, the original Thief series Hammer AIs use only a few voices with names like “ham1…”, “ham2…”, and “ham3…” so my customer Hammer character is “ham7…”  Take note of the file name.  I recently started cutting-and-pasting the file name into a text file that is published with the voice set, to make it easier for the author to program the scripts without listening to all the WAVs looking for the correct one.

You have just edited and saved one voice line.  In Sound Recorder, you will have to reload your original voice recording and edit the next line from it.  In Audacity, you simply zoom all the way out (the rightmost zoom button is “Fit Project in Window”) and find the next line.  Since the line you just saved is still highlighted, you can hit the delete key to temporarily remove it, or use the play button to find your next line and the mouse to drag-select it.  If you do not save your master recording, the removal of each line as you finish it just gets it off the screen and out of your way.  Repeat the above process of zooming in to select the exact portion of the sound you want and exporting the selection as a WAV file in the folder you created for that character.  Be careful to use a different file name for each sound. 

When you have all the lines saved, the editing for that conversation or voice set is done.  You can keep the original recording (very long, with pauses and recording artifacts) in your archives if you want, but it will not be part of what you publish.

Publishing:

Publishing your voice work means organizing your voice recordings and getting them to those who can use them.

I recommend putting each character’s lines in its own folder.  For conversations, you could also put all the lines for a single conversation in one folder.  File names should fit the rules for the game, for example Thief series WAV file names look like “sg1a0_1.wav” with 2-4 characters to identify the AI, 1-4 for the situation, and 1-2 to identify the specific instance of that situation.  File names should also be unique, so pick AI identifiers that are not already used in the game. 

I also recommend putting a text file into the folder explaining who you are, your requirements for using the voice in game mods (usually a request that you be credited and informed of the use), and enough information for the game author to understand your naming scheme and therefore use the appropriate sounds in the appropriate situations.

Put the folder(s) into an appropriately-named compressed file.  In Windows, this is as simple as right-clicking the folder and selecting Send To .. Compressed Folder which creates a ZIP file.  Finally, hang the ZIP file on a web site and tell the author where to find it, or email it directly to the author if his/her email can handle large files.  For general-use voice sets, I host the files on my web site

For voice sets that are created for specific conversations, I upload the file to my web site but do not link it from my publicly-accessible site, and send a direct link to the author.  This way, only the author can find the file, and any secrets or spoilers contained in the conversation are not released.

Final note:  practice makes perfect, so if you do not feel ready to try voice acting, try using your computer to record sound clips from other sources and then practice editing them in Audacity.  Many online web sites offer quotes from popular movies and TV shows, which are often used for system events by those wishing to personalize their system.  Pull those up in Audacity and you may be surprised at how much better you can edit them to sound through the techniques above.

Happy voice acting!

Don

brethren has written 23 articles

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>