15.ai: Difference between revisions

Browse history interactively ← Previous edit Next edit →Content deleted Content addedVisual WikitextInline

Revision as of 08:47, 17 December 2024 editSirfurboy (talk \| contribs)Extended confirmed users21,629 edits →Development: Unsourced, irrelevant. It is gone. There is no need for a "still gone" running commentary.← Previous edit		Revision as of 01:47, 19 December 2024 edit undoGregariousMadness (talk \| contribs)Extended confirmed users1,326 edits -- Draft creation using the WP:Article wizard --Tag: Disambiguation links addedNext edit →
Line 1:		Line 1:
			{{AfC submission\|t\|\|ts=20241219014703\|u=GregariousMadness\|ns=118\|demo=}}<!-- Important, do not remove this line before article has been created. -->
	{{Short description\|Real-time text-to-speech tool using artificial intelligence}}
	{{pp\|small=yes}}
	{{Use mdy dates\|date=July 2022}}
	{{Infobox website		{{Infobox website
	\| name = 15.ai		\| name = 15.ai
Line 10:		Line 8:
	\| commercial = No		\| commercial = No
	\| registration = None		\| registration = None
	\| launch_date = ~~'''Initial release''':~~ {{start date and age\|2020\|03\|12}}~~<br/>'''Last stable release''': v24.2.1~~		\| launch_date = {{start date and age\|2020\|03\|12}}
	\| type = ], ], ], ]		\| type = ], ]
			\| website = {{URL\|https://15.ai}}
	\| language = English		\| language = English
	}}		}}
			'''15.ai''' was a free ] that used ] to generate ] voices of fictional characters from ].<ref name="udn">{{cite web
	{{Artificial intelligence}}
			\|url= https://game.udn.com/game/story/10453/5189551
	'''15.ai''' was a free to use ] ] that generated ] voices from fictional characters from various media sources.<ref name="kotaku">{{cite web
			\|title= 這個AI語音可以模仿《傳送門》GLaDOS講出任何對白！連《Undertale》都可以學
	\|url= https://kotaku.com/this-website-lets-you-make-glados-say-whatever-you-want-1846062835
			\|first= 遊戲角落
	\|title= Website Lets You Make GLaDOS Say Whatever You Want
	\|last= ~~Zwiezen~~		\|last=遊戲
	\|~~first~~= ~~Zack~~		\|date= 2021-01-20
			\|website= ]
			\|access-date= 2024-12-18
			\|quote=
			\|url-status=live
			}}</ref><ref name="denfami">{{cite web
			\|url= https://news.denfaminicogamer.jp/news/210118f
			\|title= 『Portal』のGLaDOSや『UNDERTALE』のサンズがテキストを読み上げてくれる。文章に込められた感情まで再現することを目指すサービス「15.ai」が話題に
			\|last= Yoshiyuki
			\|first= Furushima
	\|date= 2021-01-18		\|date= 2021-01-18
	\|website= ]		\|website= ]
	\|access-date= ~~2021~~-01-18		\|access-date= 2024-12-18
	\|quote=		\|quote=
	\|archive-date= 2021-01-17		\|archive-date= 2021-01-18
	\|archive-url= https://web.archive.org/web/~~20210117164748~~/https://~~kotaku~~.~~com~~/~~this-website-lets-you-make-glados-say-whatever-you-want-1846062835~~		\|archive-url= https://web.archive.org/web/20210118051321/https://news.denfaminicogamer.jp/news/210118f
	\|url-status= live		\|url-status= live
			}}</ref> The application allowed users to make characters from various media speak custom text with emotional inflections.<ref>{{cite web
	}}</ref><ref name="gameinformer">{{cite magazine
			\|url= https://automaton-media.com/articles/newsjp/20210119-149494/
			\|title= ゲームキャラ音声読み上げソフト「15.ai」公開中。『Undertale』や『Portal』のキャラに好きなセリフを言ってもらえる
			\|last= Kurosawa
			\|first= Yuki
			\|date= 2021-01-19
			\|website= ]
			\|access-date= 2024-12-18
			\|quote=
			\|archive-date= 2021-01-19
			\|archive-url= https://web.archive.org/web/20210119103031/https://automaton-media.com/articles/newsjp/20210119-149494/
			\|url-status= live
			}}</ref><ref name="gi">{{cite magazine
	\|url= https://www.gameinformer.com/gamer-culture/2021/01/18/make-portals-glados-and-other-beloved-characters-say-the-weirdest-things		\|url= https://www.gameinformer.com/gamer-culture/2021/01/18/make-portals-glados-and-other-beloved-characters-say-the-weirdest-things
	\|title= Make Portal's GLaDOS And Other Beloved Characters Say The Weirdest Things With This App		\|title= Make Portal's GLaDOS And Other Beloved Characters Say The Weirdest Things With This App
Line 34:		Line 54:
	\|date= 2021-01-18		\|date= 2021-01-18
	\|magazine= ]		\|magazine= ]
	\|access-date= ~~2021~~-01-18		\|access-date= 2024-12-18
	\|quote=		\|quote=
	\|archive-date= 2021-01-18		\|archive-date= 2021-01-18
	\|archive-url= https://web.archive.org/web/20210118175543/https://www.gameinformer.com/gamer-culture/2021/01/18/make-portals-glados-and-other-beloved-characters-say-the-weirdest-things		\|archive-url= https://web.archive.org/web/20210118175543/https://www.gameinformer.com/gamer-culture/2021/01/18/make-portals-glados-and-other-beloved-characters-say-the-weirdest-things
	\|url-status= dead		\|url-status= dead
	}}</ref><ref name="~~pcgamer~~">{{cite web		}}</ref><ref name="pcg">{{cite web
	\|url= https://www.pcgamer.com/make-the-cast-of-tf2-recite-old-memes-with-this-ai-text-to-speech-tool		\|url= https://www.pcgamer.com/make-the-cast-of-tf2-recite-old-memes-with-this-ai-text-to-speech-tool
	\|title= Make the cast of TF2 recite old memes with this AI text-to-speech tool		\|title= Make the cast of TF2 recite old memes with this AI text-to-speech tool
Line 46:		Line 66:
	\|date= 2021-01-19		\|date= 2021-01-19
	\|website= ]		\|website= ]
	\|access-date= ~~2021~~-01-19		\|access-date= 2024-12-18
	\|quote=		\|quote=
	\|archive-date= 2021-01-19		\|archive-date= 2021-01-19
	\|archive-url= https://web.archive.org/web/20210119133726/https://www.pcgamer.com/make-the-cast-of-tf2-recite-old-memes-with-this-ai-text-to-speech-tool/		\|archive-url= https://web.archive.org/web/20210119133726/https://www.pcgamer.com/make-the-cast-of-tf2-recite-old-memes-with-this-ai-text-to-speech-tool/
	\|url-status= live		\|url-status= live
	}}</ref><ref name="~~rockpapershotgun~~">{{cite web		}}</ref><ref name="rps">{{cite web
	\|url= https://www.rockpapershotgun.com/2021/01/18/put-words-in-game-characters-mouths-with-this-fascinating-text-to-speech-tool/		\|url= https://www.rockpapershotgun.com/2021/01/18/put-words-in-game-characters-mouths-with-this-fascinating-text-to-speech-tool/
	\|title= Put words in game characters' mouths with this fascinating text to speech tool		\|title= Put words in game characters' mouths with this fascinating text to speech tool
Line 58:		Line 78:
	\|date= 2021-01-18		\|date= 2021-01-18
	\|website= ]		\|website= ]
	\|access-date= ~~2021~~-01-18		\|access-date= 2024-12-18
	\|quote=		\|quote=
	\|archive-date= 2021-01-18		\|archive-date= 2021-01-18
	\|archive-url= https://web.archive.org/web/20210118213308/https://www.rockpapershotgun.com/2021/01/18/put-words-in-game-characters-mouths-with-this-fascinating-text-to-speech-tool/		\|archive-url= https://web.archive.org/web/20210118213308/https://www.rockpapershotgun.com/2021/01/18/put-words-in-game-characters-mouths-with-this-fascinating-text-to-speech-tool/
	\|url-status= live		\|url-status= live
			}}</ref>
	}}</ref> Created by a ]ous developer under the alias '''15''',<ref name="automaton">{{cite web

	\|url= https://automaton-media.com/articles/newsjp/20210119-149494/
			15.ai is credited as the first example to popularize AI voice cloning (]) in ] and ].<ref>{{cite web
	\|title= ゲームキャラ音声読み上げソフト「15.ai」公開中。『Undertale』や『Portal』のキャラに好きなセリフを言ってもらえる
			\|url=https://analyticsindiamag.com/ai-origins-evolution/deepfakes-are-elevating-meme-culture-but-at-what-cost/
	\|last= Kurosawa
			\|title=Deepfakes Are Elevating Meme Culture, But At What Cost?
	\|first= Yuki
			\|last=VK
	\|date= 2021-01-19
			\|first=Anirudh
	\|website= ]
	\|~~access-~~date= ~~2021~~-01-19		\|date=2023-03-18
			\|website=Analytics India Magazine
			\|access-date=2024-12-18
			\|url-status=live
			}}</ref><ref>
			{{cite web
			\|url=https://www.inverse.com/gaming/youtube-ai-presidential-gaming-debates
			\|title=Why Biden, Trump, and Obama Arguing Over Video Games Is YouTube's New Obsession
			\|last=Wright
			\|first=Steven
			\|date=2023-03-21
			\|website=]
			\|access-date=2024-12-18
			\|url-status=live
			}}</ref> Initially launched in early 2020,<ref name="thebatch">
			{{cite web \|last=Ng \|first=Andrew \|date=2020-04-01 \|title=Voice Cloning for the Masses \|url=https://blog.deeplearning.ai/blog/the-batch-ai-against-coronavirus-datasets-voice-cloning-for-the-masses-finding-unexploded-bombs-seeing-see-through-objects-optimizing-training-parameters \|url-status=dead \|archive-url=https://web.archive.org/web/20200807111844/https://blog.deeplearning.ai/blog/the-batch-ai-against-coronavirus-datasets-voice-cloning-for-the-masses-finding-unexploded-bombs-seeing-see-through-objects-optimizing-training-parameters \|archive-date=2020-08-07 \|access-date=2024-12-18 \|website=] \|quote=}}
			</ref> the application went ] in 2021 on social media platforms like ] and ] and quickly became popular among Internet fandoms, including the '']'', '']'', and '']'' fandoms.<ref name="kotaku">{{cite web
			\|url= https://kotaku.com/this-website-lets-you-make-glados-say-whatever-you-want-1846062835
			\|title= Website Lets You Make GLaDOS Say Whatever You Want
			\|last= Zwiezen
			\|first= Zack
			\|date= 2021-01-18
			\|website= ]
			\|access-date= 2024-12-18
	\|quote=		\|quote=
	\|archive-date= 2021-01-19		\|archive-date= 2021-01-17
	\|archive-url= https://web.archive.org/web/~~20210119103031~~/https://~~automaton-media~~.com/~~articles/newsjp/20210119~~-~~149494/~~		\|archive-url= https://web.archive.org/web/20210117164748/https://kotaku.com/this-website-lets-you-make-glados-say-whatever-you-want-1846062835
	\|url-status= live		\|url-status= live
	}}</ref><ref name="~~elevenlabs~~">{{cite web		}}</ref><ref name="gamersky">{{cite web
	\|url=https://~~elevenlabs~~.io/~~blog~~/~~15-ai~~		\|url=https://www.gamersky.com/news/202101/1355887.shtml
			\|title=这个网站可用AI生成语音让ACG角色“说”出你输入的文本
	\|title=15.AI: Everything You Need to Know & Best Alternatives
			\|date=2021-01-18
	\|website=]
			\|website=]
	\|date=2024-02-07
	\|access-date=2024-11-18		\|access-date=2024-12-18
	\|url-status=live		\|quote=
			\|url-status=live
			}}</ref>
	\|archive-date=July 15, 2024
	\|archive-url=https://web.archive.org/web/20240715151316/https://elevenlabs.io/blog/15-ai
	}}</ref><ref name="resemble">{{cite web
	\|url=https://www.resemble.ai/free-15ai-character-voice-cloning-alternatives/
	\|title=Free 15.ai Character Voice Cloning and Alternatives
	\|website=Resemble.ai
	\|date= October 17, 2024
	\|access-date= 2024-11-18
	}}</ref><ref name="play.ht">{{cite web
	\|url=https://play.ht/blog/15-ai/
	\|title=Everything You Need to Know About 15.ai: The AI Voice Generator
	\|website=Play.ht
	\|date=2024-09-12
	\|access-date=2024-11-18
	}}</ref> the project used a combination of ] algorithms, ] ], and ] models to generate emotive character voices.<ref name="hashdork">{{cite web
	\|url=https://hashdork.com/15-ai/
	\|title=15.ai – Natural and Emotional Text-to-Speech Using Neural Networks
	\|website=Hashdork
	\|date=2024-05-15
	\|access-date=2024-11-18
	\|url-status=live
	\|archive-date=July 4, 2024
	\|archive-url=https://web.archive.org/web/20240704144415/https://hashdork.com/15-ai/
	}}</ref><ref name="thelinuxcode">{{cite web
	\|url=https://thelinuxcode.com/what-15ai-and-how-does-work/
	\|title=Demystifying 15.ai: How AI Generates Ultra-Realistic Text-to-Speech Voices
	\|website=TheLinuxCode
	\|date=2023-12-27
	\|access-date=2024-11-18
	\|url-status=live
	\|archive-date=December 27, 2023
	\|archive-url=https://web.archive.org/web/20231227222306/https://thelinuxcode.com/what-15ai-and-how-does-work/
	}}</ref>

			Various commercial alternatives to 15.ai appeared in the following years.<ref name="elevenlabs">{{cite web
	In early 2020, 15.ai appeared online as a ] of the ] of ] and ].<ref name="play.ht"/><ref name="thebatch">
			\|url=https://elevenlabs.io/blog/15-ai
	{{cite web \|last=Ng \|first=Andrew \|date=2020-04-01 \|title=Voice Cloning for the Masses \|url=https://blog.deeplearning.ai/blog/the-batch-ai-against-coronavirus-datasets-voice-cloning-for-the-masses-finding-unexploded-bombs-seeing-see-through-objects-optimizing-training-parameters \|url-status=dead \|archive-url=https://web.archive.org/web/20200807111844/https://blog.deeplearning.ai/blog/the-batch-ai-against-coronavirus-datasets-voice-cloning-for-the-masses-finding-unexploded-bombs-seeing-see-through-objects-optimizing-training-parameters \|archive-date=2020-08-07 \|access-date=2020-04-05 \|website=DeepLearning.AI \|quote=}}
			\|title=15.AI: Everything You Need to Know & Best Alternatives
	</ref> Its gratis nature, ease of use without ], and improvements over existing text-to-speech implementations made it popular.<ref name="gameinformer"/><ref name="kotaku" /><ref name="pcgamer" /> Some critics and ]s questioned the ] and ] of making such technology so readily accessible.<ref name="wccftech">{{cite web \|last=Lopez \|first=Ule \|date=2022-01-16 \|title=Troy Baker-backed NFT firm admits using voice lines taken from another service without permission \|url=https://wccftech.com/voiceverse-nft-service-uses-stolen-technology-from-15ai/ \|url-status=live \|archive-url=https://web.archive.org/web/20220116194519/https://wccftech.com/voiceverse-nft-service-uses-stolen-technology-from-15ai/ \|archive-date=2022-01-16 \|access-date=2022-06-07 \|website=Wccftech}}</ref>
			\|date= 2024-02-07

			\|website=]
	The site was embraced by Internet ]s such as ], '']'', and '']''.<ref name="automaton"/><ref name="Denfaminicogamer">{{cite web
			\|access-date=2024-12-18
	\|url= https://news.denfaminicogamer.jp/news/210118f
			\|quote=
	\|title= 『Portal』のGLaDOSや『UNDERTALE』のサンズがテキストを読み上げてくれる。文章に込められた感情まで再現することを目指すサービス「15.ai」が話題に
			\|url-status=live
	\|last= Yoshiyuki
			}}</ref><ref name="playht">{{cite web
	\|first= Furushima
			\|url= https://play.ht/blog/15-ai/
	\|date= 2021-01-18
			\|title= Everything You Need to Know About 15.ai: The AI Voice Generator
	\|website= Denfaminicogamer
	\|~~access-~~date= ~~2021~~-01-18		\|date= 2024-09-12
			\|website= Play.ht
			\|access-date= 2024-12-18
	\|quote=		\|quote=
	\|archive-date= 2021-01-18
	\|archive-url= https://web.archive.org/web/20210118051321/https://news.denfaminicogamer.jp/news/210118f
	\|url-status= live		\|url-status= live
			}}</ref> In January 2022, the company Voiceverse NFT, which had partnered with voice actor ], plagiarized 15.ai's work as part of their platform.<ref name="nme">{{cite web
	}}</ref><ref name="play.ht"/>

	Several commercial alternatives appeared in the following years.<ref name="elevenlabs"/><ref name="resemble"/> In January 2022, the company Voiceverse NFT ] 15.ai's work as part of their platform.<ref name="nme">{{cite web
	\|url= https://www.nme.com/news/gaming-news/voiceverse-nft-admits-to-taking-voice-lines-from-non-commercial-service-3140663		\|url= https://www.nme.com/news/gaming-news/voiceverse-nft-admits-to-taking-voice-lines-from-non-commercial-service-3140663
	\|title= Voiceverse NFT admits to taking voice lines from non-commercial service		\|title= Voiceverse NFT admits to taking voice lines from non-commercial service
Line 141:		Line 151:
	\|date= 2022-01-18		\|date= 2022-01-18
	\|website= ]		\|website= ]
	\|access-date= ~~2022~~-01-18		\|access-date= 2024-12-18
	\|quote=		\|quote=
	\|archive-date= 2022-01-18		\|archive-date= 2022-01-18
Line 153:		Line 163:
	\|date= 2022-01-17		\|date= 2022-01-17
	\|website= Stevivor		\|website= Stevivor
	\|access-date= ~~2022~~-01-17		\|access-date= 2024-12-18
	\|quote=		\|quote=
	\|archive-date= 2022-01-17		\|archive-date= 2022-01-17
Line 160:		Line 170:
	}}</ref>		}}</ref>

			In September 2022, 15.ai was taken offline due to legal issues surrounding ].<ref name="twitter">{{cite web
	The ethical implications of ] (also known as '']'') in ] led to a re-evaluation of the service by the developer, with concerns being raised regarding copyright and the unauthorized use of character voices.<ref name="play.ht"/> In September 2022, a year after its last stable release, 15.ai was taken offline.<ref name="elevenlabs"/>
			\|url=https://x.com/fifteenai/status/1865439846744871044

			\|title=The past and future of 15.ai
	== Features ==
			\|website=]
	], known for his sinister robotic voice, was one of the available characters on 15.ai.<ref name="kotaku"/>]]
			}}</ref><ref name="bnl">{{cite web
	The platform required no ] or ] to generate voices.<ref name="LaPS4"/><ref name="yahoofin"/><ref name="resemble"/><ref name="play.ht"/> Users could generate speech by entering text and selecting a character voice (optionally specifying an emotional contextualizer and/or phonetic transcriptions), with the system producing three variations of the audio with different emotional deliveries.<ref name="hashdork"/> The platform operated completely ], though the developer reported spending thousands of dollars monthly to maintain the service.<ref name="play.ht"/>
			\|url= https://businessnewsledger.com/researcher-behind-15ai-reveals-development-history-of-influential-voice-platform/

			\|title= Researcher Behind 15.ai Reveals Development History of Influential Voice Platform
	Available characters included ] and ] from '']'', characters from '']'', ] and other ] from '']'', ], ] and ] from '']'', the ], ] from '']'', the Narrator from '']'', ] from '']'', ], Dan from '']'', and ] from '']''.<ref name="LaPS4">{{cite web
			\|last= Squire
	\|url= https://www.laps4.com/noticias/descubre-15-ai-un-sitio-web-en-el-que-podras-hacer-que-glados-diga-lo-que-quieras/
			\|first= Esperanza
	\|title= Descubre 15.AI, un sitio web en el que podrás hacer que GlaDOS diga lo que quieras
	\|~~last~~= ~~Villalobos~~		\|date= 2024-12-11
	\|~~first~~= ~~José~~		\|website= Stevivor
	\|date= ~~2021~~-01-18		\|access-date= 2024-12-18
	\|website= ]
	\|access-date= 2021-01-18
	\|quote=		\|quote=
			\|website=Business News Ledger
	\|archive-date= 2021-01-18
	\|archive-url= https://web.archive.org/web/20210118172043/https://www.laps4.com/noticias/descubre-15-ai-un-sitio-web-en-el-que-podras-hacer-que-glados-diga-lo-que-quieras/
	\|url-status= live
	}}</ref><ref name="yahoofin">{{cite web
	\|url= https://es-us.finanzas.yahoo.com/noticias/15-ai-sitio-te-permite-152000712.html
	\|title= 15.ai, el sitio que te permite usar voces de personajes populares para que digan lo que quieras
	\|last= Moto
	\|first= Eugenio
	\|date= 2021-01-20
	\|website= ]
	\|access-date= 2021-01-20
	\|quote=
	\|archive-date= 2022-03-08
	\|archive-url= https://web.archive.org/web/20220308230836/https://es-us.finanzas.yahoo.com/noticias/15-ai-sitio-te-permite-152000712.html
	\|url-status= live		\|url-status= live
	}}</ref>		}}</ref>
			== Features ==

			The platform operated without requiring user registration or accounts. Users generated speech by inputting text and selecting a character voice, with optional parameters for emotional contextualizers and phonetic transcriptions. Each request produced three audio variations with distinct emotional deliveries.<ref name="tds">
			{{cite web
			\|url=https://towardsdatascience.com/generate-your-favourite-characters-voice-lines-using-machine-learning-c0939270c0c6
			\|title=Generate Your Favourite Characters’ Voice Lines using Machine Learning
			\|last=Chandraseta
			\|first=Rionaldi
			\|date=2021-01-21
			\|website=Towards Data Science
			\|access-date=2024-12-18
			\|url-status=live
			}}</ref>

			Characters available on 15.ai included ] and ] from '']'', characters from '']'', ] and other characters from '']'', ], ] from '']'', the ], and ] from '']''.<ref name="kotaku"/><ref name="pcg"/><ref name="rps"/><ref name="gi"/>
	The ] nature of the ] model ensured that each generation would have slightly different intonations, similar to multiple takes from a ].<ref name="hashdork"/><ref name="automaton"/> The application supported manually altering the ] of a generated line using ''emotional contextualizers'' (a term coined by this project), a sentence or phrase conveying the emotion of the take that serves as a guide for the model during inference.<ref name="automaton"/><ref name="Denfaminicogamer"/>

	Emotional contextualizers were representations of the emotional content of a sentence deduced via ] ] ] using ], a deep neural network ] algorithm developed by the ] in 2017.<ref>{{cite book \|last=Felbo \|first=Bjarke \|arxiv=1708.00524 \|title=Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing\|chapter=Using millions of emoji occurrences to learn any-domain representations for detecting sentiment, emotion and sarcasm \|date=2017 \|pages=1615–1625 \|doi=10.18653/v1/D17-1169 \|s2cid=2493033 }}</ref><ref>{{cite web
			The deep learning model's nondeterministic properties produced variations in speech output, creating different intonations with each generation, similar to how ]s produce different takes. The system introduced ''"emotional contextualizers,"'' which allowed users to specify the emotional tone of generated speech through guiding phrases.<ref name="tds"/> The emotional contextualizer functionality utilized DeepMoji, a sentiment analysis neural network developed at the ]. Introduced in 2017, DeepMoji processed ] embeddings from 1.2 billion Twitter posts (2013-2017) to analyze emotional content. Testing showed the system could identify emotional elements, including sarcasm, more accurately than human evaluators.<ref name="techreview">{{cite web
	\|url= https://www.theregister.com/2017/08/07/sarcasm_detector_bot_mit/
	\|title= A sarcasm detector bot? That sounds absolutely brilliant. Definitely
	\|last= Corfield
	\|first= Gareth
	\|date= 2017-08-07
	\|website= ]
	\|access-date= 2022-06-02
	\|archive-date= 2022-06-02
	\|archive-url= https://web.archive.org/web/20220602215737/https://www.theregister.com/2017/08/07/sarcasm_detector_bot_mit/
	\|url-status= live
	}}</ref> DeepMoji was trained on 1.2 billion emoji occurrences in ] data from 2013 to 2017, and outperformed human subjects in correctly identifying sarcasm in Tweets and other online modes of communication.<ref>{{cite web
	\|url= https://www.technologyreview.com/2017/08/03/105566/an-algorithm-trained-on-emoji-knows-when-youre-being-sarcastic-on-twitter/		\|url= https://www.technologyreview.com/2017/08/03/105566/an-algorithm-trained-on-emoji-knows-when-youre-being-sarcastic-on-twitter/
	\|title= An Algorithm Trained on Emoji Knows When You're Being Sarcastic on Twitter		\|title= An Algorithm Trained on Emoji Knows When You're Being Sarcastic on Twitter
Line 211:		Line 209:
	\|date= 2017-08-03		\|date= 2017-08-03
	\|website= ]		\|website= ]
	\|access-date= ~~2022~~-06-02		\|access-date= 2024-12-18
	\|archive-date= 2022-06-02		\|archive-date= 2022-06-02
	\|archive-url= https://web.archive.org/web/20220602215737/https://www.technologyreview.com/2017/08/03/105566/an-algorithm-trained-on-emoji-knows-when-youre-being-sarcastic-on-twitter/		\|archive-url= https://web.archive.org/web/20220602215737/https://www.technologyreview.com/2017/08/03/105566/an-algorithm-trained-on-emoji-knows-when-youre-being-sarcastic-on-twitter/
	\|url-status= live
	}}</ref><ref>{{cite web
	\|url= https://www.bbc.com/news/technology-40850171
	\|title= Emojis help software spot emotion and sarcasm
	\|last=
	\|first=
	\|date= 2017-08-07
	\|website= ]
	\|access-date= 2022-06-02
	\|archive-date= 2022-06-02
	\|archive-url= https://web.archive.org/web/20220602215735/https://www.bbc.com/news/technology-40850171
	\|url-status= live
	}}</ref><ref>{{cite web
	\|url= https://www.newsweek.com/emoji-computer-sarcasm-emotion-training-hate-speech-647474
	\|title= Emoji-Filled Mean Tweets Help Scientists Create Sarcasm-Detecting Bot That Could Uncover Hate Speech
	\|last= Lowe
	\|first= Josh
	\|date= 2017-08-07
	\|website= ]
	\|access-date= 2022-06-02
	\|archive-date= 2022-06-02
	\|archive-url= https://web.archive.org/web/20220602215735/https://www.newsweek.com/emoji-computer-sarcasm-emotion-training-hate-speech-647474
	\|url-status= live		\|url-status= live
	}}</ref>		}}</ref>

			== References ==
	15.ai used a ''multi-speaker model''—hundreds of voices were trained concurrently rather than sequentially, decreasing the required training time and enabling the model to learn and generalize shared emotional context, even for voices with no exposure to that context.<ref name="arxivmello">{{cite arXiv \|last=Valle \|first=Rafael \|eprint=1910.11997 \|title=Mellotron: Multispeaker expressive voice synthesis by conditioning on rhythm, pitch and global style tokens \|class=eess \|date=2020 }}</ref> Consequently, the characters in the application were powered by a single trained model, as opposed to multiple single-speaker models.<ref name="arxivmulti">{{cite arXiv \|last=Cooper \|first=Erica \|eprint=1910.10838 \|title=Zero-Shot Multi-Speaker Text-To-Speech with State-of-the-art Neural Speaker Embeddings \|class=eess \|date=2020 }}</ref> The ] used by 15.ai was scraped from a variety of Internet sources, including ], ], the ], ], ], and ]. Pronunciations of unfamiliar words were automatically deduced using ]s learned by the deep learning model.<ref name="automaton"/>

	The application supported a simplified phonetic transcription known as ], to correct mispronunciations and account for ]—words that are spelled the same but are pronounced differently (such as the word ''read'', which can be pronounced as either {{IPAc-en\|ˈ\|r\|ɛ\|d}} or {{IPAc-en\|ˈ\|r\|iː\|d}} depending on its ]). It followed the ]'s ARPABET conventions.<ref name="automaton" />
	{{clear}}

	== Background ==
	=== Artificial intelligence in speech synthesis ===
	{{Main\|Deep learning speech synthesis}}
	{{See also\|Audio deepfake}}
	]'s ].<ref name="deepmind" />]]
	In 2016, with the proposal of ]'s ], deep-learning-based models for speech synthesis began to gain popularity as a method of modeling waveforms and generating high-fidelity human-like speech.<ref name="arxiv1">{{cite arXiv \|last=Hsu \|first=Wei-Ning \|eprint=1810.07217 \|title=Hierarchical Generative Modeling for Controllable Speech Synthesis \|class=cs.CL \|date=2018 }}</ref><ref name="arxiv2">{{cite arXiv \|last=Habib \|first=Raza \|eprint=1910.01709 \|title=Semi-Supervised Generative Modeling for Controllable Speech Synthesis \|class=cs.CL \|date=2019 }}</ref><ref name="deepmind">{{cite web\|url=https://www.deepmind.com/blog/high-fidelity-speech-synthesis-with-wavenet\|title=High-fidelity speech synthesis with WaveNet\|last1=van den Oord\|first1=Aäron\|last2=Li\|first2=Yazhe\|last3=Babuschkin\|first3=Igor\|date=2017-11-12\|website=]\|access-date=2022-06-05\|archive-date=2022-06-18\|archive-url=https://web.archive.org/web/20220618205838/https://www.deepmind.com/blog/high-fidelity-speech-synthesis-with-wavenet\|url-status=live}}</ref> Tacotron2, a neural network architecture for speech synthesis developed by ], was published in 2018 and required tens of hours of audio data to produce intelligible speech; when trained on 2 hours of speech, the model was able to produce intelligible speech with mediocre quality, and when trained on 36 minutes of speech, the model was unable to produce intelligible speech.<ref name="tacotron">{{cite web\|url=https://google.github.io/tacotron/publications/semisupervised/index.html\|title=Audio samples from "Semi-Supervised Training for Improving Data Efficiency in End-to-End Speech Synthesis"\|date=2018-08-30\|access-date=2022-06-05\|archive-date=2020-11-11\|archive-url=https://web.archive.org/web/20201111222714/https://google.github.io/tacotron/publications/semisupervised/index.html\|url-status=live}}</ref><ref name="arxiv3">{{cite arXiv \|eprint=1712.05884 \|title=Natural TTS Synthesis by Conditioning WaveNet on Mel-Spectrogram Predictions \|class=cs.CL \|date=2018 \|last1=Shen \|first1=Jonathan \|last2=Pang \|first2=Ruoming \|last3=Weiss \|first3=Ron J. \|last4=Schuster \|first4=Mike \|last5=Jaitly \|first5=Navdeep \|last6=Yang \|first6=Zongheng \|last7=Chen \|first7=Zhifeng \|last8=Zhang \|first8=Yu \|last9=Wang \|first9=Yuxuan \|last10=Skerry-Ryan \|first10=RJ \|last11=Saurous \|first11=Rif A. \|last12=Agiomyrgiannakis \|first12=Yannis \|last13=Wu \|first13=Yonghui }}</ref>

	For years, reducing the amount of data required to train a realistic high-quality text-to-speech model has been a primary goal of scientific researchers in the field of deep learning speech synthesis.<ref>{{cite arXiv \|last=Chung \|first=Yu-An \|eprint=1808.10128 \|title=Semi-Supervised Training for Improving Data Efficiency in End-to-End Speech Synthesis \|class=cs.CL \|date=2018 }}</ref><ref>{{cite arXiv \|last=Ren \|first=Yi \|eprint=1905.06791 \|title=Almost Unsupervised Text to Speech and Automatic Speech Recognition \|class=cs.CL \|date=2019 }}</ref> The developer of 15.ai claims that as little as 15 seconds of data is sufficient to clone a voice up to human standards, a significant reduction in the amount of data required.<ref name="eurogamer">{{cite web \|last=Phillips \|first=Tom \|date=2022-01-17 \|title=Troy Baker-backed NFT firm admits using voice lines taken from another service without permission \|url=https://www.eurogamer.net/articles/2022-01-17-troy-baker-backed-nft-firm-admits-using-voice-lines-taken-from-another-service-without-permission \|url-status=live \|archive-url=https://web.archive.org/web/20220117164033/https://www.eurogamer.net/articles/2022-01-17-troy-baker-backed-nft-firm-admits-using-voice-lines-taken-from-another-service-without-permission \|archive-date=2022-01-17 \|access-date=2022-01-17 \|website=] \|quote=}}</ref>

	== Development ==
	15.ai was designed and created by an anonymous research scientist known by the alias ''15''.<ref name="automaton"/><ref name="elevenlabs"/><ref name="resemble"/> In his blog '']'', economist ] cited the developer of 15.ai as an example of underrated talent in AI.<ref name="marginalrevolution">{{cite web
	\|url= https://marginalrevolution.com/marginalrevolution/2022/05/the-most-underrated-talent-in-ai.html
	\|title= The most underrated talent in AI?
	\|last= Cowen
	\|first= Tyler
	\|date= 2022-05-12
	\|website= ]
	\|access-date= 2024-11-27
	\|url-status= live
	\|archive-date= 2022-06-19
	\|archive-url= https://web.archive.org/web/20220619203626/https://marginalrevolution.com/marginalrevolution/2022/05/the-most-underrated-talent-in-ai.html
	}}</ref>

	Developing and running 15.ai cost several thousands of dollars per month, initially funded by the developer's personal finances after a successful ] exit.<ref name="play.ht"/> The algorithm used by the project was dubbed '''DeepThroat.'''<ref name="play.ht"/>The project and algorithm were conceived as part of MIT's ], and had been in development since 2018.<ref name="thebatch"/><ref name="automaton"/><ref>{{cite web
	\|url=https://www.byteside.com/2021/01/15-ai-deepmoji-glados-spongebob-characters-ai-text-to-speech/
	\|title=Make GLaDOS, SpongeBob and other friends say what you want with this AI text-to-speech tool
	\|last=Button
	\|first=Chris
	\|date=2021-01-19
	\|website=Byteside
	\|access-date=2024-11-18
	\|url-status=live
	\|archive-date=June 25, 2024
	\|archive-url=https://web.archive.org/web/20240625180514/https://www.byteside.com/2021/01/15-ai-deepmoji-glados-spongebob-characters-ai-text-to-speech/
	}}</ref> The model used by 15.ai was inspired by a 2019 paper that introduced ] to text-to-speech models.<ref name="thebatch"/><ref>{{cite book \|last=Jia \|first=Ye \|arxiv=1806.04558 \|title=1806.04558\|date=2019 }}</ref>

	]'s /mlp/ board has been integral to the development of 15.ai.<ref name="gwern">{{cite journal \|last=Branwen \|first=Gwern \|date=2020-03-06 \|title="15.ai"⁠, 15, Pony Preservation Project \|url=https://www.gwern.net/docs/ai/music/index#15-project-2020-section \|url-status=live \|publisher=Gwern \|archive-url=https://web.archive.org/web/20220318160737/https://www.gwern.net/docs/ai/music/index#15-project-2020-section \|archive-date=2022-03-18 \|access-date=2022-06-17 \|website=Gwern.net}}</ref>]]
	The developer also worked closely with the Pony Preservation Project from /mlp/, the '']'' ] of ].<ref name="play.ht"/> This project was a "collaborative effort by /mlp/ to build and curate pony datasets" with the aim of creating applications in artificial intelligence.<ref>{{cite web
	\|url= https://www.equestriadaily.com/2020/03/neat-pony-preservation-project-using.html
	\|title= Neat "Pony Preservation Project" Using Neural Networks to Create Pony Voices
	\|last= Scotellaro
	\|first= Shaun
	\|date= 2020-03-14
	\|website= ]
	\|access-date= 2022-06-11
	\|archive-date= 2021-06-23
	\|url-status= live
	\|archive-url= https://web.archive.org/web/20210623210048/https://www.equestriadaily.com/2020/03/neat-pony-preservation-project-using.html
	}}</ref><ref name="ppp">
	{{cite web
	\|url= https://desuarchive.org/mlp/thread/38204261/
	\|title= Pony Preservation Project (Thread 108)
	\|last=
	\|first=
	\|date= 2022-02-20
	\|website= ]
	\|publisher= Desuarchive
	\|access-date= 2022-02-20
	\|quote= }}</ref> The ''Friendship Is Magic'' voices on 15.ai were trained on a large dataset ]d by the project: audio and dialogue from the show and related media<ref name="play.ht"/>—including ], ], ], ], and various other content voiced by the same voice actors—were ], ], and ] to remove background noise.

	The first public release of 15.ai was unveiled in March 2020, with the service experiencing intermittent availability as the developer conducted ongoing ] work.{{citation needed\|date=December 2024}} The tool gained heavy attention in ] in early 2021, with multiple gaming news outlets covering its capabilities.<ref name="pcgamer"/><ref name="kotaku"/><ref name="gameinformer"/> 15.ai saw further attention in 2022 when it was discovered that the Voiceverse NFT had used outputs from the tool.<ref name="nme"/>

	== Reception ==
	15.ai was met with a largely positive reception from users and ]. Liana Ruppert of '']'' described it as "simplistically brilliant"<ref name="gameinformer"/> and José Villalobos of '']'' wrote that it "works as easy as it looks."<ref name="LaPS4"/>{{efn\|Translated from original quote written in Spanish: ''"La dirección es 15.AI y funciona tan fácil como parece."''<ref name="LaPS4"/>}} Lauren Morton of '']'' called the tool "fascinating,"<ref name="rockpapershotgun"/> and Yuki Kurosawa of '']'' deemed it "revolutionary."<ref name="automaton"/>{{efn\|Translated from original quote written in Japanese: ''"しかし15.aiが画期的なのは「データが30秒しかない文字でも、ほぼ100％の発音精度を達成できること」そして「ごくわずかなデータのみを使って、自然な感情のこもった音声を数百以上生成できること」だという。"''<ref name="automaton"/>}} Users praised the ability to easily create audio of popular characters that sound believable to those unaware they had been synthesized. Zack Zwiezen of '']'' reported that " girlfriend was convinced it was a new voice line from ]' voice actor, ]".<ref name="kotaku"/> Natalie Clayton of '']'' wrote that "]' shrill, nasally voice works shockingly well".

	The website's impact extended beyond English-speaking media. Yoshiyuki Furushima of '']'' wrote that "it's amazing that are all synthetically generated", and Eugenio Moto of '']'' reported that "while the results are already exceptional, they can certainly get better."

	== In popular culture ==
	=== Fandom content creation ===
	<!-- Deleted image removed: ] -->
	15.ai was frequently used for ] in various ]s, including the ], the '']'' fandom, the '']'' fandom, and the '']'' fandom, with numerous videos and projects containing speech from 15.ai having gone ].<ref name="kotaku" /><ref name="gameinformer" /> The platform is credited as the impetus behind the popularization of AI voice cloning in content creation, demonstrating the potential for accessible, high-quality voice synthesis technology.<ref name="play.ht"/>

	The ''My Little Pony: Friendship Is Magic'' fandom saw a resurgence in video and musical content creation as a result, inspiring a new genre of fan-created content assisted by artificial intelligence. Some ]s weren adapted into fully voiced "episodes": ''The Tax Breaks'' is a 17-minute long animated video rendition of a fan-written story published in 2014 that uses voices generated from 15.ai with ] and ], emulating the episodic style of the early seasons of ''Friendship Is Magic''.<ref name="taxbreaks">{{cite web
	\|url= https://www.equestriadaily.com/2022/05/full-simple-animated-episode-tax-breaks.html
	\|title= Full Simple Animated Episode – The Tax Breaks (Twilight)
	\|last= Scotellaro
	\|first= Shaun
	\|date= 2022-05-15
	\|website= ]
	\|access-date= 2022-05-28
	\|quote=
	\|archive-date= 2022-05-21
	\|url-status= live
	\|archive-url= https://web.archive.org/web/20220521132423/https://www.equestriadaily.com/2022/05/full-simple-animated-episode-tax-breaks.html
	}}</ref><ref>{{Cite web \|date=27 April 2014 \|title=The Terribly Taxing Tribulations of Twilight Sparkle \|url=https://www.fimfiction.net/story/185725 \|url-status=live \|archive-url=https://web.archive.org/web/20220630170105/https://www.fimfiction.net/story/185725 \|archive-date=30 June 2022 \|access-date=28 April 2022 \|website=Fimfiction.net}}</ref>

	Viral videos from the ''Team Fortress 2'' fandom featuring voices from 15.ai include ''Spy is a ]'' (which gained over 3 million views on YouTube across multiple videos<ref group="yt">{{cite web\|url=https://www.youtube.com/watch?v=TAmhr6Was3E\|title=SPY IS A FURRY\|work=]\|date=January 17, 2021 \|access-date=June 14, 2022\|archive-date=June 13, 2022\|archive-url=https://web.archive.org/web/20220613094918/https://www.youtube.com/watch?v=TAmhr6Was3E\|url-status=live}}</ref><ref group="yt">{{cite web\|url=https://www.youtube.com/watch?v=lwQn7ISVV_8\|title=Spy is a Furry Animated\|work=]\|access-date=June 14, 2022\|archive-date=June 14, 2022\|archive-url=https://web.archive.org/web/20220614203255/https://www.youtube.com/watch?v=lwQn7ISVV_8\|url-status=live}}</ref><ref group="yt">{{cite web\|url=https://www.youtube.com/watch?v=r0FLyW86owo\|title= – Spy's Confession – \|work=]\|date=January 15, 2021 \|access-date=June 14, 2022\|archive-date=June 30, 2022\|archive-url=https://web.archive.org/web/20220630170113/https://www.youtube.com/watch?v=r0FLyW86owo\|url-status=live}}</ref>) and ''The RED Bread Bank'', both of which inspired ] animated video renditions.<ref name="automaton"/> Other fandoms used voices from 15.ai to produce viral videos. {{As of\|July 2022}}, the viral video ''] Struggles'' (with voices from ''Friendship Is Magic'') had over 5.5 million views on YouTube;<ref group="yt">{{cite web\|url=https://www.youtube.com/watch?v=UPE3vnLY3TE\|title=Among Us Struggles\|work=]\|date=September 21, 2020 \|access-date=July 15, 2022}}</ref> ], ], and ] streamers also used 15.ai for their videos, such as FitMC's video on the history of ]—one of the oldest running '']'' servers—and datpon3's TikTok video featuring the main characters of ''Friendship Is Magic'', which have 1.4 million and 510 thousand views, respectively.<ref group="yt">{{cite web\|url=https://www.youtube.com/watch?v=1V1O2gTdqHw\|title=The UPDATED 2b2t Timeline (2010–2020)\|work=]\|date=March 14, 2020 \|access-date=June 14, 2022\|archive-date=June 1, 2022\|archive-url=https://web.archive.org/web/20220601085855/https://www.youtube.com/watch?v=1V1O2gTdqHw\|url-status=live}}</ref><ref group="tt">{{cite web\|url=https://www.tiktok.com/@datpon3/video/6813618431217241350\|title=She said " 👹 "\|work=]\|access-date=July 15, 2022\|archive-date=February 21, 2022\|archive-url=https://web.archive.org/web/20220221225053/https://www.tiktok.com/@datpon3/video/6813618431217241350\|url-status=live}}</ref>

	Some users created AI ]s using 15.ai and external voice control software. One user on Twitter created a personal desktop assistant inspired by ] using 15.ai-generated dialogue in tandem with voice control system VoiceAttack.<ref name="automaton"/><ref name="Denfaminicogamer"/>

	== See also ==
	{{div col}}
	*]
	*]
	*]
	*]
	*]
	*]
	*]
	*]
	*]
	{{div col end}}

	==Notes==
	{{notelist}}

	==References==
	;Notes
	{{reflist}}		{{reflist}}
	;YouTube (referenced for view counts and usage of 15.ai only)
	{{reflist\|group=yt\|35em}}
	;TikTok
	{{reflist\|group=tt\|35em}}

	==External links==
	* {{Twitter \| id= fifteenai \| name= 15 }}

	{{Differentiable computing}}
	{{Speech synthesis}}
	{{My Little Pony: Friendship Is Magic}}

	]
	]
	]
	]
	]
	]
	]
	]
	]
	]
	]

Revision as of 01:47, 19 December 2024

This article, 15.ai, has recently been created via the Articles for creation process. Please check to see if the reviewer has accidentally left this template after accepting the draft and take appropriate action as necessary.

Reviewer tools: Inform author

15.ai
Type of site	Artificial intelligence, speech synthesis
Available in	English
Founder(s)	15
URL	15.ai
Commercial	No
Registration	None
Launched	March 12, 2020; 4 years ago (2020-03-12)

15.ai was a free web application that used artificial intelligence to generate text-to-speech voices of fictional characters from popular media. The application allowed users to make characters from various media speak custom text with emotional inflections.

15.ai is credited as the first example to popularize AI voice cloning (audio deepfakes) in memes and content creation. Initially launched in early 2020, the application went viral in 2021 on social media platforms like YouTube and Twitter and quickly became popular among Internet fandoms, including the My Little Pony: Friendship Is Magic, Team Fortress 2, and SpongeBob SquarePants fandoms.

Various commercial alternatives to 15.ai appeared in the following years. In January 2022, the company Voiceverse NFT, which had partnered with voice actor Troy Baker, plagiarized 15.ai's work as part of their platform.

In September 2022, 15.ai was taken offline due to legal issues surrounding artificial intelligence and copyright.

Features

The platform operated without requiring user registration or accounts. Users generated speech by inputting text and selecting a character voice, with optional parameters for emotional contextualizers and phonetic transcriptions. Each request produced three audio variations with distinct emotional deliveries.

Characters available on 15.ai included GLaDOS and Wheatley from Portal, characters from Team Fortress 2, Twilight Sparkle and other characters from My Little Pony: Friendship Is Magic, SpongeBob, Sans from Undertale, the Tenth Doctor Who, and HAL 9000 from 2001: A Space Odyssey.

The deep learning model's nondeterministic properties produced variations in speech output, creating different intonations with each generation, similar to how voice actors produce different takes. The system introduced "emotional contextualizers," which allowed users to specify the emotional tone of generated speech through guiding phrases. The emotional contextualizer functionality utilized DeepMoji, a sentiment analysis neural network developed at the MIT Media Lab. Introduced in 2017, DeepMoji processed emoji embeddings from 1.2 billion Twitter posts (2013-2017) to analyze emotional content. Testing showed the system could identify emotional elements, including sarcasm, more accurately than human evaluators.

References

遊戲, 遊戲角落 (2021-01-20). "這個AI語音可以模仿《傳送門》GLaDOS講出任何對白！連《Undertale》都可以學". United Daily News. Retrieved 2024-12-18.{{cite web}}: CS1 maint: url-status (link)
Yoshiyuki, Furushima (2021-01-18). "『Portal』のGLaDOSや『UNDERTALE』のサンズがテキストを読み上げてくれる。文章に込められた感情まで再現することを目指すサービス「15.ai」が話題に". Denfaminicogamer. Archived from the original on 2021-01-18. Retrieved 2024-12-18.
Kurosawa, Yuki (2021-01-19). "ゲームキャラ音声読み上げソフト「15.ai」公開中。『Undertale』や『Portal』のキャラに好きなセリフを言ってもらえる". AUTOMATON. Archived from the original on 2021-01-19. Retrieved 2024-12-18.
^ Ruppert, Liana (2021-01-18). "Make Portal's GLaDOS And Other Beloved Characters Say The Weirdest Things With This App". Game Informer. Archived from the original on 2021-01-18. Retrieved 2024-12-18.
^ Clayton, Natalie (2021-01-19). "Make the cast of TF2 recite old memes with this AI text-to-speech tool". PC Gamer. Archived from the original on 2021-01-19. Retrieved 2024-12-18.
^ Morton, Lauren (2021-01-18). "Put words in game characters' mouths with this fascinating text to speech tool". Rock, Paper, Shotgun. Archived from the original on 2021-01-18. Retrieved 2024-12-18.
VK, Anirudh (2023-03-18). "Deepfakes Are Elevating Meme Culture, But At What Cost?". Analytics India Magazine. Retrieved 2024-12-18.{{cite web}}: CS1 maint: url-status (link)
Wright, Steven (2023-03-21). "Why Biden, Trump, and Obama Arguing Over Video Games Is YouTube's New Obsession". Inverse. Retrieved 2024-12-18.{{cite web}}: CS1 maint: url-status (link)
Ng, Andrew (2020-04-01). "Voice Cloning for the Masses". DeepLearning.AI. Archived from the original on 2020-08-07. Retrieved 2024-12-18.
^ Zwiezen, Zack (2021-01-18). "Website Lets You Make GLaDOS Say Whatever You Want". Kotaku. Archived from the original on 2021-01-17. Retrieved 2024-12-18.
"这个网站可用AI生成语音让ACG角色"说"出你输入的文本". GamerSky. 2021-01-18. Retrieved 2024-12-18.{{cite web}}: CS1 maint: url-status (link)
"15.AI: Everything You Need to Know & Best Alternatives". ElevenLabs. 2024-02-07. Retrieved 2024-12-18.{{cite web}}: CS1 maint: url-status (link)
"Everything You Need to Know About 15.ai: The AI Voice Generator". Play.ht. 2024-09-12. Retrieved 2024-12-18.{{cite web}}: CS1 maint: url-status (link)
Williams, Demi (2022-01-18). "Voiceverse NFT admits to taking voice lines from non-commercial service". NME. Archived from the original on 2022-01-18. Retrieved 2024-12-18.
Wright, Steve (2022-01-17). "Troy Baker-backed NFT company admits to using content without permission". Stevivor. Archived from the original on 2022-01-17. Retrieved 2024-12-18.
"The past and future of 15.ai". Twitter.
Squire, Esperanza (2024-12-11). "Researcher Behind 15.ai Reveals Development History of Influential Voice Platform". Business News Ledger. Retrieved 2024-12-18.{{cite web}}: CS1 maint: url-status (link)
^ Chandraseta, Rionaldi (2021-01-21). "Generate Your Favourite Characters' Voice Lines using Machine Learning". Towards Data Science. Retrieved 2024-12-18.{{cite web}}: CS1 maint: url-status (link)
"An Algorithm Trained on Emoji Knows When You're Being Sarcastic on Twitter". MIT Technology Review. 2017-08-03. Archived from the original on 2022-06-02. Retrieved 2024-12-18.

Categories: