15.ai, or "How To Make GLaDOS Say the F Word"


“This is a text-to-speech tool that you can use to generate 44.1 kHz voices of various characters. The voices are generated in real time using multiple audio synthesis algorithms and customized deep neural networks trained on very little available data (between 15 and 120 minutes of clean dialogue for each character). This project demonstrates a significant reduction in the amount of audio required to realistically clone voices while retaining their affective prosodies.”

at peak hours, e.g. right now, the servers may be clogged, making the voice lines take a while to be generated.

and as promised in the title, here is GLaDOS saying the F word, as well as a few other samples from other characters;

(source: angry gamers around the world)

(source: Genesis 1:2)

[add sample of Narrator quoting Moby Dick later]

[add sample of Wheatley telling a terrible yo mama joke later]


Thanks very much for this! :innocent:

I mean, 'cause there’s Twi’s voice :yum:

1 Like


I just tried Narrator from the Stanley Parable, it sounds echoy so some reasons


AI Replica

I tried a longer voice line and this is the result

Another AI Replica

1 Like

15.ai reads Problem Sleuth

15.ai is fucking fun.

Fun fact: GLaDOS is worried for me after calling me a “weirdo with a Canberra image”

Probably the AI trying to make the words fit together.

Come on no Portal 2 Announcer?

probably not enough audio

Updates for April 13th, 2020:

Engorde. e vai tomar no c*

I can confirm, TF2 is unselectable, however 15.ai-generated speech for the Solider sounded off when i tested it before his death.

the website has been down for 8 hours. not a good sign :frowning:

Well probably because they’re improving voices :wink:

no, normally that stuff happens behind the scenes. it shouldn’t take 11.5 hours



clear body fk discourse

1 Like

would probably say more than “brb” if that was the case


Update for April 17th, 2020:

  • Voices being retrained: website unusable for an unknown amount of time
1 Like

@appledoo, I’m kinda right already :slight_smile:

Sample of Twilight Sparkle’s voice after retraining; compare to my sample in this thread

Perfect! And thanks very much! :yum: