Trying to recreate GLaDOS voice with Text-To-Speech software - Games

This has been done before, many times, in fact, there was a webpage that offered an automated service for GLaDOS-like voice, the problem was, this voice was far from similar to GLaDOS. So what am I trying to do? Well, I got hold (by means I won't tell you yet) of a copy of the Loquendo TTS suite, what this is is probably the most advanced and realistic TTS voice program nowadays, It has advanced options for emphasis, lenght of words, word stress, tone of voice, prominence and pronunciation. companies all over the world use it for costumer service and the like, here are some demos of how good the voice is: [url]http://dl.dropbox.com/u/25318683/Susan.wav[/url] [url]http://dl.dropbox.com/u/25318683/Allison.wav[/url] and that, paired with this tutorial ([url]http://www.youtube.com/watch?v=1oQn66gvwKA[/url]) and some others around the net, I'm trying to get an almost perfect Glados voice. So far I've only made 4 tests, and they don't really sound a lot like GLaDOS, just a bit, I've also run down some issues: the voice I'm currently using has big troubles with H's, words like though, shark, thing get a bit fucked up. Also, the software has big time issues with windows 7 and I have to make a few workarounds to save the voice as audio. The software I'm using to modify the voice is "Melodyne" a voice modifying software (for singers, like Rebecca Black) that lets me add a variety of effects and modify tones and pitch. Here's what I've got so far (each one is better than the last one): 1st test: [url]http://dl.dropbox.com/u/25318683/addsd.melo.wav[/url] 2nd test: [url]http://dl.dropbox.com/u/25318683/chamber.melo.wav[/url] (this one fucks up a bit in the beginning) 3rd test: [url]http://dl.dropbox.com/u/25318683/pace.melo.wav[/url] 4th test (I think it's the best one so far, and the weirdest): [url]http://dl.dropbox.com/u/25318683/potato.melo.wav[/url] 5th test: [url]http://dl.dropbox.com/u/25318683/didwellenough.melo.wav[/url] 6th test: [url]http://dl.dropbox.com/u/25318683/melodyne.melo.wav[/url] As you can see, I'm trying different pitches, and there are many factors that get in the way, like emotions, tones, if it's a question, an exclamation and many other things. I'll also try to use another voice. Now, I'm not an expert in sound and whatever but I'm sure some of you are or are good at it, so right now I want people to play with this and see if they can come up with an almost perfect GLaDOS voice, and I know it can be done. For now, If you want to give this a try PM or better, email me at: [email]andrescanelones10@gmail.com[/email] and I'll give you the info. This would be great for Modders, so that they can include GLaDOS dialogue on their mods. And you know, creating OC in general.

Those actually sound quite like her. A bit more work and you'll be there! [editline]22nd May 2011[/editline] I've always wanted to hear GLaDOS say "I am coming". [sp]4th test[/sp] [editline]22nd May 2011[/editline] How long does it take to make each sound clip? Can you just run an automated set of commands that does it for you or do you have to do each one manually?

[QUOTE=barttool;29982692]4th test (I think it's the best one so far, and the weirdest): [url]http://dl.dropbox.com/u/25318683/potato.melo.wav[/url] [/QUOTE] [b]THIS JUST IN:[/b] GLaDOS has a potato fetish.

That's fantastic. Great job.

4th test sounds very good actually

[QUOTE=aero1444;29982745]Those actually sound quite like her. A bit more work and you'll be there! [editline]22nd May 2011[/editline] I've always wanted to hear GLaDOS say "I am coming". [sp]4th test[/sp] [editline]22nd May 2011[/editline] How long does it take to make each sound clip? Can you just run an automated set of commands that does it for you or do you have to do each one manually?[/QUOTE] Well, the workarounds are not always needed, the suite comes with an advanced TTS editor (the one that has issues) and a simple one (that does work), when I don't need to add emphasis or any especial command, I just use the simple TTS editor. Otherwise, I set the recording device to the Stereo Mix, and opened the sound recorder, and then played the voice in the TTS editor. doesn't take long but it nags me a lot.

[QUOTE=PelPix123;29982853]I put about four solid months of work into this a bit back. Valve uses a voice from AT&T to generate the prompts for Ellen to imitate. They put the lines into the TTS and then have her imitate what comes out. This is entirely void in Portal 2, where she totally drops the formal, robotic air, but it's what they did for Portal 1. Her inflection depends on the mood she's in, of which there are way too many to name. Here are two examples: [media]http://soundcloud.com/kylebloss/mood1-1[/media] [media]http://soundcloud.com/kylebloss/mood2-1[/media] And that's the result of spending four months studying how a character from a video game talks :smithicide: I love being a sound designer. It's a great profession. And an original line with a try at using scripting to create sarcasm. You just lose some much humanity with TTS voices. It just isn't worth it. That's why I gave up. [media]http://soundcloud.com/kylebloss/testtrack-5[/media] A lot of the quirks in the way GLaDOS speaks were created by this TTS, like the pitch that she pronounces "well" in most of her lines. This TTS was used for the original concept of GLaDOS before she was ever a character, back when they were testing level design and trying to make the physics system actually work.[/QUOTE] If i ever make a Portal 2 mod i want you in my team.

I'd suggest slowing down the things she says manually in some type of a sound editor, because the tts is playing words next to each other a bit to fast in my opinion. Add a small gap between words.

will you disclose your method of obtaining it in a pm? i can't even find a "buy this fucking software" button on their website.

[url]http://glados.biringa.com/[/url] Is quite good.

This can really help us mappers to add atmosphere into our maps! Thanks for figuring this out!

[QUOTE=PelPix123;29982853]I put about four solid months of work into this a bit back. Valve uses a voice from AT&T to generate the prompts for Ellen to imitate. They put the lines into the TTS and then have her imitate what comes out. This is entirely void in Portal 2, where she totally drops the formal, robotic air, but it's what they did for Portal 1. Her inflection depends on the mood she's in, of which there are way too many to name. Here are two examples: [media]http://soundcloud.com/kylebloss/mood1-1[/media] [media]http://soundcloud.com/kylebloss/mood2-1[/media] And that's the result of spending four months studying how a character from a video game talks :smithicide: I love being a sound designer. It's a great profession. And an original line with a try at using scripting to create sarcasm. You just lose some much humanity with TTS voices. It just isn't worth it. That's why I gave up. [media]http://soundcloud.com/kylebloss/testtrack-5[/media] A lot of the quirks in the way GLaDOS speaks were created by this TTS, like the pitch that she pronounces "well" in most of her lines. This TTS was used for the original concept of GLaDOS before she was ever a character, back when they were testing level design and trying to make the physics system actually work.[/QUOTE] Amazing work! Well, TTS does have a lot of issues, and never really recreates the complete feeling of GLaDOS' voice, but we're not shooting for the stars here anyways, I just want a similar enough GLaDOS voice, enough to seem like GLaDOS and not some regular robotic voice, and I think we can come close to that. [editline]22nd May 2011[/editline] [QUOTE=Shounic;29986784]will you disclose your method of obtaining it in a pm? i can't even find a "buy this fucking software" button on their website.[/QUOTE] email me.

You done good, OP. Valve also used Melodyne for GLaDOS's official voice lines.

I've made some more tests, getting a bit better at using melodyne, and it turns out, the program is really great for shaping tones and pronunciation, I can add emphasis, lenght and stress on each word, giving the sense of emotion. Although, I'm not perfect at it yet, so sometimes, words can sound weird. here's what I got: [url]http://dl.dropbox.com/u/25318683/didwellenough.melo.wav[/url] [url]http://dl.dropbox.com/u/25318683/melodyne.melo.wav[/url] If you speak spanish, spell this in english: T N S L P P B N T S O [url]http://dl.dropbox.com/u/25318683/pipibientieso.melo.wav[/url] Once again, if you want to give it a shot and see if you can make it better, email me @ [email]andrescanelones10@gmail.com[/email] and I'll give you the info.

[QUOTE=barttool;29989687]I've made some more tests, getting a bit better at using melodyne, and it turns out, the program is really great for shaping tones and pronunciation, I can add emphasis, lenght and stress on each word, giving the sense of emotion. Although, I'm not perfect at it yet, so sometimes, words can sound weird. here's what I got: [url]http://dl.dropbox.com/u/25318683/didwellenough.melo.wav[/url] [url]http://dl.dropbox.com/u/25318683/melodyne.melo.wav[/url] If you speak spanish, spell this in english: T N S L P P B N T S O [url]http://dl.dropbox.com/u/25318683/pipibientieso.melo.wav[/url] Once again, if you want to give it a shot and see if you can make it better, email me @ [email]andrescanelones10@gmail.com[/email] and I'll give you the info.[/QUOTE] The 2nd one was good but the first one not so much

This is awesome. The other day I was thinking about how we should have [I]some[/I] type of standardized Text-To-Speech voice for use in making Portal 2 mods and one based on GLaDOS seems like the best bet. If it's impossible we could just make a new voice and claim it's a secondary A.I.

Bravo!

Needs to be less nasally and a little deeper. Thats all I can say.

No one can replace Ellen McLain in my eyes.

[QUOTE=toxicspill3457;29992997]No one can replace Ellen McLain in my eyes.[/QUOTE] perhaps in your ears?

Remember, the way it's done is to read 'kind of' like the text to speech. Completely remove pitch modulation, up the formant by one and a half and optionally snap all the pieces to their nearest pitch. That's how the in-game commentary said it was done.

[QUOTE=toxicspill3457;29992997]No one can replace Ellen McLain in my eyes.[/QUOTE] I guess you haven't seen this yet then (32 seconds) [media]http://www.youtube.com/watch?v=0reihQtBTSE[/media]

[QUOTE=Baboo00;29997215]I guess you haven't seen this yet then (32 seconds) [media]http://www.youtube.com/watch?v=0reihQtBTSE[/media][/QUOTE] Slightly noticeable difference, but damn that's accurate.

[QUOTE=Smashman;29986813][url]http://glados.biringa.com/[/url] Is quite good.[/QUOTE] Message added to queue successfully. Position in queue: 11,173.

[QUOTE=Cronos Dage;29997655]Message added to queue successfully. Position in queue: 11,173.[/QUOTE] I really doubt it's that number, I entered one hours ago and it came with that exact same number, which is damn unlikely.

Yeah, it's Portal 1-esque so probably very few use it now.

[QUOTE=CakeMaster7;29997420]Slightly noticeable difference, but damn that's accurate.[/QUOTE] Most of the diference comes from the way she speaks, and her tone of voice beign slightly higher than McLain's.

these all seriously sound like shit. just get an actual real-life girl to record it. it's not hard.

it sounds as if Glados had a daughter

[QUOTE=DOG-GY;30000715]these all seriously sound like shit. just get an actual real-life girl to record it. it's not hard.[/QUOTE] Dude, the whole point of this is to avoid needing a person to record the voices, of course having a girl record her voice is way better, but this thread is called "trying to recreate GLaDOS voice with text-to-speech software".