Re: the funnest site i have ever found
Re: the funnest site i have ever found If ya check out This Fifteen AI Update Video, all will be heard.You hear the original scene from where the quote's based first, followed by the older Fifteen AI and newer AI examples respectively after the 2 minute mark. The first two minutes are like, tiny examples, with long segment music transitions.NB. I didn't make said video.PS. The first voice is Durpy. URL: https://forum.audiogames.net/post/598747/#p598747 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found If ya check out This Fifteen AI Update Video, all will be heard.You hear the original scene from where the quote's based first, followed by the older Fifteen AI and newer AI examples respectively after the 2 minute mark. The first two minutes are like, tiny examples, with long segment music transitions.NB. I didn't make said video. URL: https://forum.audiogames.net/post/598747/#p598747 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found I can't find that to. It's frustrating, cause I wanna listen to them. URL: https://forum.audiogames.net/post/598718/#p598718 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found Has anyone figured out how to play the examples on 15.ai's website? Can't find anything that'll do it. URL: https://forum.audiogames.net/post/598711/#p598711 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
2020-11-30
Thread
AudioGames . net Forum — Off-topic room : ammericandad2005 via Audiogames-reflector
Re: the funnest site i have ever found here are a fiew voices I'd like to have included in a future update (I wonder how to contact the guy to suggest them):James rolf (AVGN)Bill O'rilyAlex Trebec (I have ripped voice clips from the official Jeopardy! video game for wii for voice training purpuses)John Guilbert (Jeopardy! announcer)Berl ivesJimmy DeranteFred AstairGreer garsenMario (as portrayed by: charles martonnette, Walker boon, Captain Lu Albano, Peter cullin, Harrie Shore and bob hoskins) URL: https://forum.audiogames.net/post/594321/#p594321 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found oh theirs a link their to do it if you really want URL: https://forum.audiogames.net/post/594295/#p594295 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found This site is indeed fun to mess with. Spent at least an hour yesterday having fun with it, which is rare for me.Some of the voices have reverb on them, probably because they were taken from public address speeches etc. But a large number of them seem clean, even stuff I wouldn't expect such as the Simpsons. I think Gilbert Gottfried has done some studio voice work so maybe that's why he sounds clean, and some of the game voices could've been trained from ripped voice clips from games (Snake and the boss for instance). Speaking of Metal Gear I'd love to see a Kernel Camble! But others are still cleaner than I expected, so it makes me wonder where the voice clips for training were sourced from.FYI, Gilbert Gottfried's voice in particular cracks me up as it is, so hearing him randomly emphasize certain words gave me a good laugh.Would love to be able to mess with this tech offline... take random voice clips of myself and see if I can construct a synthetic voice of it for instance. Of course one has to be very careful when distributing tools to do this because they can be misused. But the tech is downright exciting, and just illustrates why machine learning is such an intriguing thing for me. I know nothing about how it works, but every time I see a machine learning algo related to audio, I become excited to hear it in action, and especially excited to try it. URL: https://forum.audiogames.net/post/594291/#p594291 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found This site is indeed fun to mess with. Spent at least an hour yesterday having fun with it, which is rare for me.While some of the voices have reverb on them which is understandable, others seem clean. The Simpsons for instance, Gilbert Gottfried, and many others. I wonder where the vovoice training clips were sourced from.FYI, Gilbert Gottfried's voice in particular cracks me up as it is, so hearing him randomly emphasize certain words gave me a good laugh.Would love to be able to mess with this tech offline... take random voice clips of myself and see if I can construct a synthetic voice of it for instance. Of course one has to be very careful when distributing tools to do this because they can be misused. But the tech is downright exciting, and just illustrates why machine learning is such an intriguing thing for me. I know nothing about how it works, but every time I see a machine learning algo related to audio, I become excited to hear it in action, and especially excited to try it. URL: https://forum.audiogames.net/post/594291/#p594291 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found This sight is indeed fun to mess with. Spent at least an hour yesterday having fun with it, which is rare for me. Gilbert Gottfried's voice cracks me up as it is, so hearing him randomly emphasize certain words gave me a good laugh.Would love to be able to mess with this tech offline... take random voice clips of myself and see if I can construct a synthetic voice of it for instance. Of course one has to be very careful when distributing tools to do this because they can be misused. But the tech is downright exciting, and just illustrates why machine learning is such an intriguing thing for me. I know nothing about how it works, but every time I see a machine learning algo related to audio, I become excited to hear it in action, and especially excited to try it. URL: https://forum.audiogames.net/post/594291/#p594291 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found this is funny, and a lot of them are clean and dry, i dont know how they split all that and make it clean, an example of a clean voice is that vegeta or whatever it's called one, it's good for a frase like thisnote. people that ofend quite easily should not read this, and children should not read this too. it has swear words. hypersensitive people should not read this, because they might make them feel bad or somethingprepare for the swear wordspreparefortheswearwords!you are damaging my friend, you song of a bitch, i'll kill you, you fucking bastard, your corpse will be mine, i'll punch you so hard that i'll end your life, and you will not be able to revive you again URL: https://forum.audiogames.net/post/594265/#p594265 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found well i hope its soon because i wana try it URL: https://forum.audiogames.net/post/594264/#p594264 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found Just gotta keep watching.What I posted earlier was his latest tweets. URL: https://forum.audiogames.net/post/594245/#p594245 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found do we know when 15 ai will be back up? i am following them on twitter but nothing URL: https://forum.audiogames.net/post/594234/#p594234 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found At this time, Fifteen AI has 67PATRONS$242PER MONTHOriginally, he didn't even want to do a Patreon thing, but folks wanted a way to support him.About the quickly spinning up the voice thing, it's because Vo Codes uses a different approach that isn't as time consuming than the Fifteen AI method. As for Patreon with Vo.codes, it's probably because Fifteen AI has gotten more coverage talk than Vo Codes has sadly. URL: https://forum.audiogames.net/post/594232/#p594232 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found sad URL: https://forum.audiogames.net/post/594229/#p594229 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found I just feel bad using it as the guy is (somehow) paying 2000 dollars a month to run the server and has 0 patrons on patreon. URL: https://forum.audiogames.net/post/594225/#p594225 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found what do you mean by quickest spinning up of a voice URL: https://forum.audiogames.net/post/594178/#p594178 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found For accuracy in the voice, Fifteen AI takes the cake!However, for the quickest spinning up of a voice, Vo Codes is better in that regard. URL: https://forum.audiogames.net/post/594177/#p594177 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found I must say this is quite interesting. URL: https://forum.audiogames.net/post/594176/#p594176 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found so vo.codes is better? URL: https://forum.audiogames.net/post/594166/#p594166 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
2020-11-30
Thread
AudioGames . net Forum — Off-topic room : ammericandad2005 via Audiogames-reflector
Re: the funnest site i have ever found the main difference is actually in the voice variity. 15.ai is biest towards a certain hazbro property with serious social and political implications (I will not name which one) with very fiew exceptions and every update removes voices that most people would be interested in (I missed the oportunity to use the zander mobis SSBU announcer voice back when that was an option). vo.codes as greater variity in it's voices including former presidents (and a soon to be former president), Celebraties like Judy dench, meme fotter like DR. Phill and tommy waiso (yes, that one from "the room"), and even mainstreem video game characters like sonic, solid snake and the scout from tf2. One miner nitpick I have over them in this regard is that they chose cnn ankers over ones from fox, a decision that has serious political implications. Why they chose that faker tucker carlson over the honest Bill O'rily we will never know. URL: https://forum.audiogames.net/post/594161/#p594161 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found @5 Always keep an eye on their Twitter:http://twitter.com/fifteenai/@everyoneIf you'd like to support Fifteen AI, join them on Patreon: https://www.patreon.com/15aiAdding to what @4 said (which the below said response is taken from the Wayback Machine Snapshot of the 26th of October 2020):This is a text-to-speech tool for generating 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. This project demonstrates not only a significant reduction in the amount of audio required to realistically clone voices while retaining their affective prosodies, but also the feasibility of an on-demand, stable, and autonomously-improving speech synthesis application that aims to mimic a voice of limited availability. As of October 2020, this current iteration of the algorithm (v7.0) is the most cutting-edge in the fields of voice cloning and speech synthesis.First, a dictionary lookup table is used to dissect each English word into its respective phonemes. The dictionary has been compiled by scraping the Oxford Dictionaries API, Wiktionary, and the CMU Pronouncing Dictionary. If a word does not exist within the dictionary, the network deduces the pronunciation of the word using the phonological rules it has learned from training on the LibriTTS dataset.The perceived emotion of the entire text is deduced using DeepMoji (deepmoji.mit.edu), a sentiment analysis model "trained on 1.2 billion tweets with emojis to understand how language is used to express emotions." This embedding is then used to bias the output toward the desired emotion using global style tokens jointly trained on top of a customized multi-speaker text-to-speech synthesis setup.Using these methods, the TTS model is able to generate voices of varying degrees of emotion despite never having been exposed to emotive data of the character during training. Furthermore, multiple characters can be trained simultaneously, significantly reducing the amount of time required compared to if one were to train the character models individually.Fig. 1: "I call it: 'emoji analysis.'"Previously, a major impediment to the development of high-quality multi-speaker text-to-speech systems was the ostensible negative effect that the embeddings had on the quality of the output audio (for lack of a better description, this caused the synthesized voice to sound "crispy"). This project shows that it is indeed possible to get high-quality audio while supporting both multi-speaker embeddings and global style tokens, and further demonstrates that a voice can be convincingly cloned — emotion and all — with as little as 30 seconds of data.This algorithm has not yet been used on real-life persons or their respective likenesses. If you encounter a "deepfake video" of celebrities or politicians online that sound strangely stilted or monotonous, they are not associated with this project; they were likely created using an unrelated program. This project has not yet been published or made open-source. URL: https://forum.audiogames.net/post/594116/#p594116 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found @5 Always keep an eye on their Twitter.@everyoneAdding to what @4 said (which the below said response is taken from the Wayback Machine Snapshot of the 26th of October 2020):This is a text-to-speech tool for generating 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. This project demonstrates not only a significant reduction in the amount of audio required to realistically clone voices while retaining their affective prosodies, but also the feasibility of an on-demand, stable, and autonomously-improving speech synthesis application that aims to mimic a voice of limited availability. As of October 2020, this current iteration of the algorithm (v7.0) is the most cutting-edge in the fields of voice cloning and speech synthesis.First, a dictionary lookup table is used to dissect each English word into its respective phonemes. The dictionary has been compiled by scraping the Oxford Dictionaries API, Wiktionary, and the CMU Pronouncing Dictionary. If a word does not exist within the dictionary, the network deduces the pronunciation of the word using the phonological rules it has learned from training on the LibriTTS dataset.The perceived emotion of the entire text is deduced using DeepMoji (deepmoji.mit.edu), a sentiment analysis model "trained on 1.2 billion tweets with emojis to understand how language is used to express emotions." This embedding is then used to bias the output toward the desired emotion using global style tokens jointly trained on top of a customized multi-speaker text-to-speech synthesis setup.Using these methods, the TTS model is able to generate voices of varying degrees of emotion despite never having been exposed to emotive data of the character during training. Furthermore, multiple characters can be trained simultaneously, significantly reducing the amount of time required compared to if one were to train the character models individually.Fig. 1: "I call it: 'emoji analysis.'"Previously, a major impediment to the development of high-quality multi-speaker text-to-speech systems was the ostensible negative effect that the embeddings had on the quality of the output audio (for lack of a better description, this caused the synthesized voice to sound "crispy"). This project shows that it is indeed possible to get high-quality audio while supporting both multi-speaker embeddings and global style tokens, and further demonstrates that a voice can be convincingly cloned — emotion and all — with as little as 30 seconds of data.This algorithm has not yet been used on real-life persons or their respective likenesses. If you encounter a "deepfake video" of celebrities or politicians online that sound strangely stilted or monotonous, they are not associated with this project; they were likely created using an unrelated program. This project has not yet been published or made open-source. URL: https://forum.audiogames.net/post/594116/#p594116 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found Fifteen AI (once the web site is up), is run by a guy who works at MIT. Vo Codes is a guy from Atlanta Georgia.For demos of Fifteen AI and Vo Codes (in good and not so good/random examples), check out this play list, and for a play list exclusively for Fifteen AI demos, check out this play list instead. URL: https://forum.audiogames.net/post/594114/#p594114 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found well, 15.ai isn't working for me, only giving me their Twitter page. Do I need to use it on a certain browser or is it undergoing maintenance? URL: https://forum.audiogames.net/post/594110/#p594110 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found 15.ai has GLaDOS and in my opinnion better acuracy URL: https://forum.audiogames.net/post/594092/#p594092 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found So, what is the difference between 15 ai and vocodes? URL: https://forum.audiogames.net/post/594085/#p594085 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found Adding to the above, here's more news (via the Fifteen AI Twitter) on what's up from its sister site (not made by the same person who does the Vocal Playground). Quote Tweet HummusCrow @TheKillerCrow1 · Replying to @fifteenai Are you ok? This whole thing seems pretty taxing on your mental health. 15 @fifteenai -- I'm good, I just enjoy being sarcastic once in a while. It's almost ready to go – just need to wrap everything up neatly in a web-friendly format. Quote Tweet Jota @Jose_Madrid01 · Replying to @fifteenai LET GO little question, is Glados available in this version? if it isn't, take your time, I know it will be good. 15 @fifteenai -- There will be three functional voices from Portal/Portal 2, one of which might come as a surprise due to how little data there is for that character – but it performs just as well as the others, if not better. URL: https://forum.audiogames.net/post/594050/#p594050 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
Re: the funnest site i have ever found
Re: the funnest site i have ever found Adding to the above, here's more news (via the Fifteen AI Twitter) on what's up from its sister site (not made by the same person who does the Vocal Playground). 15 @fifteenai -- I'm good, I just enjoy being sarcastic once in a while. It's almost ready to go – just need to wrap everything up neatly in a web-friendly format. Quote Tweet HummusCrow @TheKillerCrow1 · Replying to @fifteenai Are you ok? This whole thing seems pretty taxing on your mental health. 15 @fifteenai -- There will be three functional voices from Portal/Portal 2, one of which might come as a surprise due to how little data there is for that character – but it performs just as well as the others, if not better. Quote Tweet Jota @Jose_Madrid01 · Replying to @fifteenai LET GO little question, is Glados available in this version? if it isn't, take your time, I know it will be good. URL: https://forum.audiogames.net/post/594050/#p594050 -- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
