8D I love singing vocaloid songs~
To record, go buy a mic. You can get a cheap decent-ish one for around $15. Download Audacity (free audio editing program) and get a karaoke version of the song you want to sing. Record your voice using Audacity, then mix it with the karaoke version. These are some mixing guidelines from YTC forum:
The Beginner's Guide to Mixing [Audacity]
Tired of apologizing for terrible mixing on every cover? Don't know what "clipping" and "compression" mean? Frustrated by your microphone's apparent love for buzzing noise? Want to make it a little easier on the judges' ears when you audition? Then this guide is for you!
It's a bit long, so it may be better to just read the section you're interested in.
Not meant for pros or the lazy =P You still have to think! Also not a replacement for practicing your singing >_>.
Update: July 18, 2010
Removed Specific Adjustments, Fix Clipping, Volume/Timing sections.
Added Mixing - Panning, Amp, Normalize, Gain sections.
Added Effect - Radio, Echo, Autotune sections.
Added Before Mixing - Clipping, Timing, Pitch sections
Added Examples section
"Equalization" section re-written!
Mentions other contributors in this thread =D
Broken up into three parts
This guide will cover:
Audacity, LAME, Recording/Editing
Before Mixing - Timing, Pitch, Clipping
Mixing - Noise Removal, Compression, EQ, Amplify/Normalize, Gain, Reverb, Pan
Effects - Radio, Echo, Autotune
Exporting to mp3
Audacity and LAME:
Audacity is open-source, has an excellent wiki, and solid functionality. We'll be using Audacity 1.3.12 Beta for this guide.
You'll also need LAME, in order to save mp3 on Windows. Follow these instructions.
Audacity Recording and Editing Basics:
The easy stuff:
Recording via the recording button (toolbar)
Importing your BGM (just drag and drop is fine)
Adjusting microphone sensitivity with the slider with the microphone icon (toolbar)
Zooming in and out in time (a.k.a the x axis) with the +/- magnifying glasses (toolbar)
Zooming in and out in the y axis by clicking and shift+clicking on a track's y axis labels
Selecting a piece of audio with your mouse, cutting if necessary (ctrl+x)
Selecting the whole track by clicking on the space where it says "Stereo, 44100 Hz" etc.
Changing the graphical display of a track to Waveform, Waveform (dB), etc. (that triangle button beside the name of the track)
Making the track display area bigger by dragging the bottom edge of the track. This will make the y axis labels display more information, which will be useful for mixing.
Adjusting the volume (or gain) of a track (slider on the left side of the track with -/+ signs)
Before Mixing - Timing:
Nothing says "I didn't practise enough" more than starting to sing half a second after when you were supposed to, or finishing a phrase with a syllable or two still unsung. Sure, you can blame it on the music starting too suddenly or something, but you don't see better singers doing it =P. It's possible to edit timing, but it's much easier to fix it by practicing your singing.
Recognize: Play your vocals and the original song at the same time (import them both into audacity). Make sure they start at the exact same time by zooming in in time and cutting out bits of silence at the beginning. It'll be obvious during playback which parts you messed up the timing on.
Recognize: Audacity sometimes has a infuriating habit of adding a bit of delay in front of your recording as it prepares itself to record. This can range from 40 to 400 milliseconds. Even 40 milliseconds of delay is noticeable, so it's up to you to make this right! Zooming in in time and looking at the waveform helps.
Prevent: Nothing you can do about the Audacity lag, other than getting a better computer =P. As for your own singing, suck it up and practise! =D
Before Mixing - Pitch:
You're off-tune? And you hope mixing and editing can save you? You're right, but it's hard. Very hard. Harder than five-year-old cheese. Audio engineers do it for pop idols who can't sing any better than your favourite nico singer (actually most nico singers are probably much better than the likes of Hannah Montana). But you're neither an audio engineer nor a pop idol (yet), so you'll have to do with good ol' fashioned practice...
Recognize: If you've got a good ear, you'll hear it. If you don't have a good ear, someone else will hear it, so ask. How do you know if you have a good ear? This test can tell you.
Prevent: Practice practice practice... it's hard, I know. Pay attention to the pitch you're producing, try singing a bit more slowly, watch out when you go high or low, whatever you notice you're weak in, practice it.
Before Mixing - Clipping:
All microphones have a certain level of maximum sound energy they can convert to electrical energy to send to your computer. Any difference between the energy received and the energy sent on is simply lost. As a result, the recording of such strong vocals sound like they're missing bits of signals, as if they've been "clipped" out. Clipping is best fixed by properly setting the sensitivity on your mic, and not by mixing or editing.
Recognize Clipping: Change the graphical display of your track to "Waveform". Do any of those waves touch the cieling or floor? If so, you might have a little bit of clipping. If you find that the waves are hugging the cieling or floor for seconds at a time, you've clipped, man, and you've clipped BADLY.
Prevent Clipping: Turn down the sensitivity of your mic (that slider in the toolbar, with the microphone icon next to it) and re-record until your vocals no longer have waveforms that touch the cieling/floor, especially during the loud parts. If you think this makes the soft parts too soft, I know already you're going to like the compression section of this guide =D.
Prevent Clipping: Also, make sure you're not so close to the mic you're about to devour it. If you have a regular mic, put it to the side of your mouth, instead of directly in front, to avoid "boom" sounds caused by breathing into the mic. If you have a condenser mic, consider a pop filter. You know, one of these.
Okay, so maybe you can fix clipping a little bit: Krystal doesn't like it, but *whispers* Effect -> Clip Fix... Try it out if you only have a little clipping. But be warned, it takes a REALLY long time.
Mixing - Noise Removal:
There's always a bit of background noise. No, I'm not talking about your brother's yelling downstairs. I'm talking about the hum of your computer and the ambient buzz in the air. Your brain might tune it out for you, but the microphone will not. Unfortunately it's hard for the computer to distinguish noise from voice, so with any noise removal process there comes a little distortion in vocals. The skill in mixing here comes from the right balance between noise removal and voice preservation.
Select a few seconds of the noise you want to remove, and go to Effect -> Noise Removal. Click "Get Noise Profile". This tells Audacity what noise to remove.
Now select a portion of your vocals and go to Effect -> Noise Removal again.
Noise reduction (dB): How much to reduce the noise by. More reduction means less noise, but also more voice distortion.
Other settings: Don't worry about them until you're pro enough. (Actually, I only know what they do, but not how to use them effectively. The default settings work fine though.)
• Use the preview button. You'll notice it only gives you the first few seconds of whatever piece of audio you selected. This'll help you adjust the settings until you're satisfied with the effect.
• Once you're satisfied, remember the settings and click cancel. Now select the whole track and Effect -> Noise Removal again. Enter the settings you decided on and click OK.
Mixing - Compression:
Dynamic range refers to the difference between the volume of the loudest sound and the softest sound. Raw vocals have a HUGE dynamic range, much larger than your BGM, in most cases. That's why oftentimes if your verse is just right your chorus gets too loud, or if your chorus is just right you can't hear the verses anymore. Compression "compresses" the dynamic range so the two are closer together in terms of volume, thus blending in with the BGM which has a similar dynamic range. The mixing skill here is to reduce differences in volume, but not so much that you can't hear the differences between powerful vocals and "sweet" vocals.
Change the graphical display of your vocal track to "Waveform (dB)" (remember that little triangle thing next to the track name?) and drag the bottom edge of the track until you have lots of informative labels displayed on the y axis. Don't be afraid to make the track so large as to fill the screen.
You'll notice general differences in volume between various parts of your vocals. Record approximately how loud (Audacity records the loudest as 0 dB and the softest as -60 dB) your soft parts and loud parts are. Say you found that they were -25 and -10 dB respectively.
Now, select a portion of your recording you'd like to preview, preferably containing a second of soft vocals followed by a second of loud. For example, the transition to a chorus.
Effect -> Compressor
Threshold: How loud the vocal has to be before compression is applied to it. We want to compress the loud vocals while leaving the soft vocals as they are. For our values of -25 dB soft and -10 dB loud, we'll set the threshold to -20 dB. Thus anything louder than -20 dB (such as our loud -10 dB vocals) will be compressed.
Ratio: How much the vocals to be compressed will be compressed. 2:1 means that the dynamic range of whatever that passes the threshold will be cut in half.
Other settings: They're fine as they are.
• Preview, and play with the ratio until the volumes are more equivalent between the soft and loud vocals, but not so much so that you can't hear the difference in power anymore.
• Remember the settings, cancel, select whole track, and apply the compressor with the settings you decided on.
• There's a curious trend in the music industry to heavily compress the dynamic range so as to get the loudness of every part of the song as high as possible. This makes sense, since it'll be easier to hear high and low frequencies when it's louder. And if two identical songs, one slightly louder than the other, were to be played, the louder one generally is regarded as better. Many people think heavy compression isn't good (I'm one of those people, since I like classical, and play piano. Dynamics is very important... Compressing until only 3dB remain, like TV commercials are, is unthinkable to me.), but that's the way it is right now. Wiki up "loudness war" if you're interested.
Mixing - Equalization:
Every pitch is identified via a frequency of the sound waves carrying the pitch. The higher the frequency, the higher the pitch you hear. The human voice typically ranges from 80 to 1100 Hz, with low vocals obviously at the lower end and high vocals at the higher end. Men typically have vocals centered between 80-500 Hz (not counting falsetto), and 170-1100 Hz for women (though the women's range covers more Hz, the relationship between frequency and pitch is not one-to-one. For each octave you go up in pitch, you'd have to DOUBLE the frequency; thus more frequency change is needed to go from high C to high D than going from low C to low D). The purpose of the EQ is to boost or diminish the volume of sound based on their frequency. For example, boosting 80 - 200 Hz might make your bass guitar sound more prominent.
If your vocals are drowned out by the BGM, you can make the BGM quieter. But as Ciel pointed out, you don't need to make the entire BGM quieter - just the frequencies where they interfere with your vocals' frequencies. In effect creating "space" in the BGM for your vocals. But which frequencies? Well, Audacity has a neat tool...
Select a representative part of your vocals. Say, the first verse, bridge, and chorus together. Now go to Analyze -> Plot Spectrum.
Whoa, it's a graph. Frequencies on the X axis, volume on the Y axis. So if you see a peak at 400 Hz, that means there are a lot of notes at around 400 Hz. From this graph you can get a feel for what range of frequencies you're singing in.
Don't fuss about being exact. Once you start getting comfortable with the equalizer, you'll know that ranges, not numbers, are what you'll be working with.
Now, think about what parts of the song is interfering with your vocals. Let's use Campanella as an example. It starts out simple, with hardly any BGM. But in the chorus it builds up and by the third chorus there's drums and cymbals and piano and even rocketships. My vocals rang true in the beginning but were drowned out near the end.
My strategy was to find out my vocals' frequency range (which we just did with Plot Spectrum), then make the BGM quieter in that range whenever I feel like I'm being drowned out, usually the chorus. I ended up doing the EQ with every chorus, and with harsher settings in the final chorus. But how do you work the EQ?
Select the portion of BGM you want to apply the EQ to, and go to Effect -> Equalization
Whoa, it's another graph. Y axis is volume, X axis is frequency. The default line is at 0 (no adjustment) for all frequencies. You can manipulate the line with the mouse. Try it out. For me, I made a small valley of about -5 dB from 150-600 Hz. Your spectrum might be different.
Alternatively, you can select the "Graphic EQ" radial button and have sliders instead of messing with the line yourself.
Other settings: If you're itchy, try them out. Just don't do anything permanent. If not, leave them alone lol.
• Preview doesn't do much here, since you need to hear your vocals at the same time. So to experiment here you'll need to apply the EQ, play it back, and if you're dissatisfied, you'll have to undo and do it again. Yeah, it's one of Audacity's weaknesses - but hey, it's free.
• Once you're satisfied, click "save as" and name your curve. Now you can apply the same EQ to other parts of the BGM that drown out your vocals.
Mixing - Amplification:
Every effect affects volume. Noise removal reduces the volume of whatever it recognizes as noise. Compression reduces the differences in volume. Equalization adjusts volume based on frequency. Amplify affects volume much more simply. It's a pure addition/subtraction in volume. The skill in mixing here comes from knowing where your vocals need boosting. Is a specific part too quiet? Is the whole track too loud?
Select something that needs boosting (the low notes that you lacked power in, and can thus barely hear, maybe?) and go to Effect -> Amplify.
Amplification (dB): How much louder/quieter you want it to be.
New Peak Amplitude (dB): Audacity calculates how many dB it will have to amplify your selection to make the peak (loudest part of the selection) whatever dB you entered here, then changes the "Amplification" field to reflect this calculation. End result is normalization (see next section) to whatever dB you entered. By default this is set to 0.0 dB.
Allow clipping: If you check this, you can boost volume above 0.0 dB, but clipping will result. I recommend not checking it.
• Preview, adjust, and apply. Play back to make sure you haven't made it too much louder/softer as to make it stand out too much from the rest of the vocals.
Mixing - Normalization:
Ever had an mp3 that was quieter than most others in your collection? If you were to make the song louder so that the max volume in this song is the same as the max volume in another song (typically 0 dB), you'd be normalizing it.
Normalization is useful, but there's already a way to do it with Amplify, and it's also included in the Compressor, if you remember.
Mixing - Gain:
Every track has a -/+ slider on the left side its display. If you can't see it, drag the bottom of the track to make the display area bigger. It has the same effect as Amplify, but limited to exactly one track at a time. So why bother?
It can be adjusted on the fly. Meaning you can adjust it while the music is playing. It also displays how much gain you're applying in dB. This is excellent for finding out exactly how many dB of amplification that quiet part of your vocals should get.
Once you know how much dB to amplify, you can put the gain slider back to 0 dB, and use Effect -> Amplify instead to make the actual changes.
Why not use the gain slider to make volume changes? It applies to the whole track whether you like it or not, so if you only want to make one part of the track louder, you're out of luck. Also, unlike Amplify, there is no "allow clipping" checkbox to leave unchecked, so it won't warn you if you're clipping.
Mixing - Reverberation:
Singing in the bathroom obviously sounds different from singing in your bedroom, which in turn is different from singing in a concert hall. The reason is echo and reverberation. If you compare your freshly noise-removed, compressed, and equalized vocals with the vocals from some songs, you'll notice that despite all your mixing, your vocals still sound very... naked. Very raw. But that doesn't mean you should make yourself sound like you're in the Globe Theatre. You just have to match the reverb of your BGM, or at least the reverb of the original Miku vocals or whatever. The skill in mixing here comes from being able to add reverb that is pleasing, but not readily noticeable (it'll distract listeners from your beautiful singing, you know?). As in, you'll notice if you compare, but not if you simply listen.
Select a suitable preview section. Preferably containing a second of a few words followed by a long drawn-out vowel. (Kaku yuugoro ni saaaaaaaaa~)
Effect -> GVerb (some people use Echo, but GVerb is more flexible and takes less processing time).
The settings are complicated. You should start out with the presets here. I like "The Quick Fix" for most songs.
• Preview. Try to aim for something that sounds pleasantly full, but natural at the same time. Adjust the amount of reverb by changing these settings:
• Early reflection level: Loudness of the first echoes. It's once again from -60 dB (softest) to 0 dB (loudest).
• Tail level value: Loudness of the echoes of the echoes, as they "die away". This is what makes reverb vocals sound so full and pleasant. Also from -60 dB to 0 dB.
• Remember your settings, cancel, select the whole track, and apply the settings you chose.
• You might want to hear your vocals again with the BGM, since the preview only plays the track you selected by itself. You might need to undo the GVerb and do it again with different settings.
• Sometimes a good pair of headphones can be a liability. What you hear as just the right amount of reverb someone else with just the good ol' iPod headphones might hear as, well, nothing. Sound quality also differs by sound card. You'll just have to learn about these the hard way, though! So once you export your mp3 later, test it out on another computer, or ask your friend to compare two versions, one with reverb and one without.
Mixing - Panning:
Stereo means different sounds signals can be sent to the left or the right speaker. Biologically speaking, your brain interprets a sound as coming from the left if it receives the sound from the left ear a few milliseconds faster than the right ear. Most mixing programs can create the illusion of your vocals coming from the left or right by panning.
Underneath the Gain slider on each track is an L/R slider that controls how close the sound from this track will sound to the left or right side.
Harmonies and background singing are good targets for panning to the left or right while your main vocals remain centered.
If you're mixing a duet or chorus, there are even more possibilities for panning. Be creative!
Effect - Radio:
A.K.A. the "tinny" effect, "walkie-talkie" effect, "buzzy" effect, etc. This one is actually produced with *gasp* the EQ! Theory is to cut off the high and low frequency components of your voice, but there's a convenient preset in Audacity you can make use of.
Effect -> Equalization. In Select curve:, there's "amradio". This simulates the sound from AM Radio stations, which are mostly talk, news, etc. Which makes sense, because if you look at the curve, everything other than the typical speaking range is getting cut.
You can modify this curve if you want, make the slopes sharper, move the peak to a higher frequency if your voice is higher, make the peak cover a smaller range, etc.
Use the preview button to experiment! When you're done, you can save your curve for future use.
Effect - Echo:
You can also use this for reverb, but GVerb is better for that, and Echo is more intuitive for ... well, echoing. Maybe you have a song that has the dramatic soft to powerful shifts like Starduster and Last Night, Good Night, where the choruses have a bit of echo to accentuate their contrast from the soft parts of the song. Or maybe you wanted to make the last note of the song echo just to sound cool. Just remember to use it moderation!
Select where you want to preview, and Effect -> Echo...
Delay time (s): time between each echo
Decay factor: how much the sound decays with each echo. 0 means complete decay (no echo), 0.5 means each echo is half as loud as the last one, 1 means the echo will never die out.
• Fun: to hear what going crazy sounds like, apply a delay time of 1 and a decay factor of 1 to, say, 20 seconds of your singing. =D
Effect - Autotune:
This basically picks a scale, and "rounds your pitch to the nearest note", to use a mathematical analogy. Human singing, though, is not that simple, and that's why autotuned voices generally sound very unrealistic. But maybe that's what you're going for, like in Campanella.
Audacity doesn't come with autotune, but there's a plugin we can download to do the same thing. ChoAkkar introduced it in the second page of the topic. It's called GSnap. Download it here. And extract the contents of the archive into the plugin folder (located where you installed Audacity)
Restart Audacity and you should now see GVST: GSnap... in the Effects menu.
How do you use GSnap? I don't actually know... I installed it because I thought I might use it for the Campanella audition, but I ended up leaving it pretty prestine. Maybe someone else will contribute? If not, I'll update this after I try it out.
Exporting to mp3:
Make sure to save your work often! And BTW, Audacity projects can take up a gigabyte of space if you did heavy editing and mixing, so make sure you move some old anime or something.
The LAME plugin we installed at the beginning was to allow for mp3 exporting. Go to File -> Export.
You can export any format from the list, but most people will choose mp3
When you do, click on "options" and change the quality. Unless you're stingy about 10 mb of space, use the highest quality setting (320 kbps). This simply tells you how much sound data will be put into the file to represent each second of music (kbps = kilobits per second).
After you press save, you can enter some tag information. I used to put my name in, but got embarrased after and now I just leave mine blank.
So you wanna hear what mixing sounds like? Here I'll provide mp3s that I edited so that different versions cycle through every few seconds as the song progresses. Sometimes the difference is subtle, and sometimes really obvious. Turning up the volume usually accentuates the differences, so if you can't tell, crank it up!
Persona's Campanella Cover - Raw vocals vs. Mixed vocals.
The radio effect in the beginning lalalas
Noise is noticeable in the unmixed parts whenever there's silence (But not in the middle instrumental portion cause I silenced that part, even in the unmixed version. Can you imagine how distracting the noise would be if I left it in?)
Reverb is noticeable, especially in verses
EQ on the BGM is present, but subtle, in the second chorus
EQ begins to be REALLY noticeable in the speaking part and the third chorus
Mixed version has some talking at the end in radio effect, and EQ on the BGM helped bring that out
Ciel's Nakimushi Ensemble Cover - Lowered BGM vs. EQ'd BGM vs. Mastered
First one is where BGM is on the whole quietened in order to let the vocals breathe. The second is not by making the whole BGM quiet, but only certain frequencies, using the EQ. The third is mastering, which I still don't quite get.
The mastered track is generally louder, but the vocals are unchanged, so only the instrumentals got louder. So I guess if doing EQ on the BGM to make room for the vocals is like making a hole in the ground to fit a pipe inside, then mastering is like using concrete to fill in the holes leftover and making the whole thing smooth again...? Maybe?
You'll have to ask Ciel for what else she did... As you can see I'm still fuzzy on what she means by mastering. Pros will be pros =).
That was crazy long, wasn't it? Actually, this is the second version of the guide. The first version was eaten by the forum cause I was writing this for too long and my login timed out. I clicked preview and it told me to log in. I logged in and it showed me a blank "new thread" page. I seriously started getting a headache for a few minutes there. But I found an extension in firefox that prevents this sort of thing, so it'll never happen to me again =D.
Hope it helped. More importantly, I hope people won't tl;dr this... <_< Right? >_> Right. I guess it'll help if it's stickied, but that's if you think this guide is good! =P So please! Add mixing to your arsenal and make sure the quality of your voice is not lost to nefarious plots of mics and comps.
:3 hope it helps?