[Voice Acting] Looking to improve.

[Pr0nogo]

From the CC thread we jacked:

[youtube]http://www.youtube.com/watch?v=spiHpv5wxW8[/youtube]

I'll be honest when I say I couldn't understand any more than maybe 5% of the spoken words.

I guess working with heavily edited voices for 11 years dulls ones hearing after a while. My point remains, however. I don't consider those comparable to the originals.

Eh, better luck next time then. I fucked up the last Khoranis line horribly.

I don't think it's really the speech itself. I think it's the flanges and reverb you're using. I know we talked about reverb before, and I think you reverb usage has improved since the previous examples you gave me, but it's still too "strong" if I may be so crude. This is the biggest thing that disorients my ability to understand the actual words, because I find the reverb to be more powerful than the diction.

To alleviate this issue you can play around with the mix balance in the reverb settings. I Believe Dry is the % of unedited original, and Wet is the effected variant. My Fear reverb uses 140% dry 428% wet. Also something worth playing around with would be the Smooth and Absorption settings. Again, my reverb for the Fear uses 25% smooth and 626ms absorption. The attack time is 238ms and the length is 615ms.

If you are using a compressor, which is recommended, it's best to use the reverb after the compressor as opposed to before. This is because the compressor will end up expanding the tails of the reverb and can further distortion.

The second aspect of the voices, especially with Khoranis, is the flanges are far too "warbly". This makes voices extremely hard to understand and is one of my biggest gripes with amateur editors. Laconius' voices aren't perfect, but the fact he avoids the warbly effects is one of the biggest reasons why they remain legible. I also find that, as long as I can avoid disruption/distortion due to phasers/warbles (which are a product of slower modulated effects), I can apply a ton of effects and the voice can remain legible with compression.

I use compression and dynamics processing and whatnot. I'll look into the reverb stuff but I'm fairly certain the distortion you attribute to the flange is primarily the reverb. Either way, I'll look into the flange. Any tips on reducing warble?

Well, Laconius tells me he uses deturner - I can't remember where the source of my effects originated from. The thing about the effects is the two delays and the Period vastly alter how they sound, and those are the things you should start playing with first. If you find the effect is dulling out everything too much your goal should probably be to make it "tighter", so the effect sounds more liquid/smooth and less distorted.

In Dynamics, to reduce the flattening you seem to be getting on the frequencies, you may wish to try raising the frequency in which the compressor takes effect on the rightmost tab.

I am about to be busy for a bit so I'll catch up on this again later.

Whilst I eagerly await Mesk's response(s) I will be putting his existing advice to good use. If anyone else has comments or tips, feel free to chime in. The videos that I post as examples here will be unlisted, with my final version being the only one to appear on my channel page.

Thanks in advance, everyone (especially Mesk).

[Lavarinth]

I agree with Mesk, though not so heavily. If I just listen and don't read, it's a bit more difficult to understand the voices. Reading makes them instantly clear, however.

[IskatuMesk]

I was writing a huge article a time ago in regards to voice acting but when CC went down I lost the save data in restarting firefox.

One of the things I am starting to learn how to do is to voice act normally. That is, to make voices without any effects. I am not accustomed to this - I've always voiced alien races, doing stuff without effects was only necessary for MFTGATRL and the standards for that only demanded the voices be silly. However, now I am doing a ton of non-edited voices for 2042, Black Sun, and several independent game projects that asked for some assistance.

In doing this I realized I actually did not know how to bring emotion into the voice for a character that was not a Villain and yet keep it believable. I was able to pull it off, but it was very difficult. I learned that speaking as clearly as possible, sometimes more slowly than otherwise to be expected, was necessary to really defining the diction in each word. This is why I mentioned speaking a bit more slowly - it allows you to practice shaping each individual word. Then you can try to speak faster.

Also, effects work best with words spoken clearly. Most of my effects mesh into a sea of base if I speak too fast. Every syllable must be pronounced to maximize the potential of my editing suites. If I don't do this, the voice just plainly doesn't sound any good. Although most Starcraft campaigns would never use the kind of heavy effects I find necessary to bring my Undead to life, you are still building with the same blocks.

Reverse - Reverb - Reverse - Reverb

Reverbs build valleys out of the gaps between words with their trails, and it should be your goal to provide those valleys with breathing room. Thus, slowly spoken words fair better than fast ones. The longer the tails, the more likely they are to crash into each other in fast speech. The stronger the reverbs, the higher their volume, the greater the valleys will be.

All of my initial UD used extensively longer trails, but the more I learned about how the settings were effected by frequencies the more I was able to make the reverbs shorter and shorter without requiring long reverbs to add a truly formidable psionic effect. It is perfectly reasonable to mesh reverbs with echo to achieve unique effects, but the echos must be extremely short and faint.

http://www.youtube.com/watch?v=oLyHRDB-Kkc

The key to editing is to achieve the most with the absolute least amount of effects. The best of my voices were the simpler ones, like the GoreKhan and Xul`Amon, whom I've since long lost the presets for.

I try to make sure a voice can always stand on its own without subtitles. With the Fear I struggle to achieve this, because I often speak too fast.

[Pr0nogo]

I don't reverb twice.

I'll keep your advice on diction and speaking slower in mind.

[Legion]

Every voice is different.

This has been a hobby of mine of 15 years. I'm still learning every single time I try. Good thread. I'm watching this...

Edit:

I liked your voices, Pro. They're different, but that may also be due to the mic quality.

Show me the effect some time.

[Pr0nogo]

Here's attempt #2.

[youtube]http://www.youtube.com/watch?v=j-_kP86bdx4[/youtube]

[IskatuMesk]

I listened while eating chicken (the irony will not be lost on Laconius, I'm sure), not reading any text at all.

The beginning half;

This was much better than before. I understood most of the voices, especially in the beginning. The reverbs are still quite intense but not nearly as bad as they were before. The big problem that stands out now is that all of the voices sound very similar.

The second half;

Holy shit Khoranis was understandable! For the most part. You started talking really fast again during this part and this broke not only the mood but made the difficult hard to understand again. At least the effects weren't nearly as much of a problem as they were before.

When voice acting you have to assume that most people are stupid and slow like me, and that talking really fast + adding a bunch of effects is not always easy to understand. Some of my previous voices, like the old Sethis tests, used absolutely insane edits. I could make them out very well when I listened to them on their own using my headphones at high volume, and another dude I talk to a lot who is a big fan of death metal and such also told me they were relatively easy to figure out.

However, every single other person did not like the voice, and when laced against background noise, it was not nearly as easy to figure out the words without subtitles or without knowing the script.

http://www.youtube.com/watch?v=TT0LI8A4E10

Despite all of the issues with Sethis' voice, I still find it easier to understand than your voices, despite my background fan going on and despite not reading a script or looking at subtitles. There's two reasons for this -

First of all, it's my voice, so it's embedded into my brain what he's saying. This makes it monumentally easier to understand your own edits. It is one of the most difficult things I have had to compensate for. I don't really seek feedback on my voice acting. I don't really seek feedback on much of anything. So when I make voices for something that will one day be public or in a video or something, I have to make a lot of calculated guesses and presumptions.

1 - Talking fast is harder for me to understand than talking slowly. I have to talk fast for certain charaters, so I limit their effects enormously. Sometimes, having a sharp reverb at all can really clash with fast words and it's best to have an extremely simple echo instead (like stereo whispers but shorter). However, it is always better to talk just a bit slower. I think you probably talk a lot faster than I do in general, again because I am Canadian and I am really slow in the head, and a lot of voicing and casting has kind of worn this voice into me. So, the only thing I can say here is practice, practice, practice. You said you have a girlfriend? Get her to hear the video without any subtitles or text or any context and see what she has to say.

Second of all, my listening conditions are extremely high. I have high performance 2.1 speakers and Steelseries Siberia headphones - the same ones you see sc2 progamers using. So my audio hardware produces a level of clarity I actually have never heard on someone else's computer. HKS' newer speakers are pretty decent, but his old ones weren't good for picking out frequencies due to their insane low-end. Another guy's computer I had used, I listened to my own music on it, and there was no base at ALL. Obviously, I can't just say "LET'S MAKE THE BEST EXPERIENCE FOR EVERYONE LOL" because that would be stupidity/carelessness on the level of indie devs/Blizzard. However, this is not something you should entirely ignore either.

When making a voice I don't aim for something that I can understand but is still on the fringe of my ability to comprehend. With the Fear's voice, that I posted earlier, sometimes his words clash into a sea of base. That's what I want to avoid with the Fear. With your voice, it loses legibility enormously when you talk fast and the reverb merges all of the words into a sea of bounding volume. I suspect if I put my headphones on and listened to them in my recording setup I would find they sound a lot better. But I'm intentionally not doing that so I can tell you what most people are going to probably be hearing.

What settings are you using for Compression, and when are you using it?

[Pr0nogo]

It's funny, because she did listen to it (I was over her house earlier) and she was like "I don't understand all them big words. What's a Cathedral?" I facepalm'd.

So she doesn't help much.

Anyhow, I'll work on talking slower with Khoranis (I talked fast - not because it was habit, but because it was something I thought could possibly define his behaviour/personality). Thanks for the tips, as always.

Compression - I currently use the Dynamics Processing preset "VocalLimit, -26thresh, +16dB" (I used to use "4:1, -24, Very Fast Attack").

[Legion]

Pr0nogo, after listening more closely-

Your vocals aren't loud enough and therefore, the effects applied kind of drown the voice. It sounds like you're applying layer upon layer, but this usually results in unclarity. Find one or two "smothering effects" (flanger, chorus) and then apply your reverbs or echoes. More than 2 flangers or choruses tends to really mess up the sample.

Check out my channel. My effects may sound "heavy", but's really only 2 at most. Aside from echo and reverbs

http://www.youtube.com/user/yearsofwork

[Pr0nogo]

Would another possible fix be amplifying my basic waveform after static reduction?

[Legion]

Would another possible fix be amplifying my basic waveform after static reduction?

No, loudness isn't the problem. It's clarity. You'd just end up having an amplified unclear voice.

Really, the trick is let go of smothering effects as much as you can - meaning: as much as you can while it still sounds like the protoss you want it to. It's really not easy. Don't let anyone tell you different.

[Pr0nogo]

Attempt #3. Spoke slower in general. Differentiated between character voices (although Azolus might be too tame for a Protoss). Hellion may have regressed in quality, not really sure on that one.

Also, for some reason, Sinar now reminds me of Mesk's normal voice.

[youtube]http://www.youtube.com/watch?v=bLVIUHs92L8[/youtube]

Link to the first half raw, before I edited them: http://www.sendspace.com/file/pujtnr

[Lavarinth]

Pr0nogo, how many times to you record the voice? Do you vary the performances each time? Do you retake the entire script in whole or do you repeat the lines? How much time to you spend prior placing your as the character? Do you speak as the character outside of the lines?

Immersion is key to a great performance. Just like live action acting, you must become the character, and if you fail to do that your real self will show through the performance.

[Pr0nogo]

I record it until I'm satisfied with the end result. In this case, six times for most of the characters. I vary the performances each time with my base voice (see the past attempts), and when I'm testing my voice before the actual recording process it doesn't usually take very long physically. Mentally, I have to picture the hordes of Zerg sacking my Carrier, and then suddenly switch to a Scout fighter pilot navigating past Scourges and Mutalisks, and then to a High Templar that's being slaughtered by Dark Templar, and then to that templar's murderers, etc.

I do the recording of a single character's lines in one take. Then I prepare my voice and mentality for the next character. The whole process of recording (not editing) VE's first briefing took me about twenty minutes this time. Effects took about three because I knew what I was doing.

As for outside of the lines, I sometimes prank call people pretending to be Tassadar if that counts. (seriously though, I just test my voice heavily before recording)

[Lavarinth]

Ah, and there you are. I actually enjoyed and understood this last performance much more easily. It was definitely clearer. Adding onto the fact that a Protoss is not entirely very easy to understand to begin with, I'd say these performance for each character is quite adequate and well done.