Tacotron 2
Tacotron 2 jẹ opin-si-opin ọrọ-si-ọrọ eto lati Google (2017) ti o yi ọrọ kikọ taara sinu a mel-spectrogram, eyi ti a nkankikan vocoder iyipada sinu igbesi aye ọrọ.
Akopọ
O ṣe agbejade awọn gbigbasilẹ ohun ti o ni idije eniyan lori awọn ipilẹ bọtini.
Jin Dive
Tacotron 2 ni awọn ẹya akọkọ meji. Ni akọkọ, nẹtiwọọki ti o tẹle-si-tẹle pẹlu akiyesi ka awọn kikọ ọrọ ati asọtẹlẹ fireemu mel-spectrogram nipasẹ fireemu. Ayipada koodu kan yi awọn ohun kikọ pada si awọn aṣoju ti o farapamọ, ẹrọ ifarabalẹ ipo-itọju kan ṣe deede ọrọ si awọn fireemu ohun, ati pe oluyipada autoregressive kan njade spectrogram nigba ti 'aami iduro' kan kọ ẹkọ nigbati ọrọ ba pari. Ẹlẹẹkeji, olugbohunsafẹfẹ WaveNet kan ti o yipada mel-spectrogram yẹn sinu fọọmu igbi aise. Nipa pipin iṣoro naa ni ọna yii, Tacotron 2 kọ ẹkọ prosody, pronunciation, ati pacing lati data pẹlu imọ-ẹrọ ọwọ diẹ. O ṣaṣeyọri Dimegilio ero itumọ tumọ si isunmọ awọn gbigbasilẹ alamọdaju, ṣiṣe ni ami-ilẹ kan ni iṣelọpọ ohun ti ara ati awoṣe fun TTS nkankikan nigbamii.
Imọ-imọ-ẹrọ
Mel-spectrogram jẹ wiwo onilàkaye laarin awọn nẹtiwọọki meji: o jẹ iwapọ ati irọrun fun awoṣe akiyesi lati ṣe asọtẹlẹ, sibẹsibẹ ọlọrọ to fun vocoder lati tun ohun afetigbọ giga-giga ṣe. Ifarabalẹ ifarabalẹ ipo ṣe idilọwọ awọn ikuna ti o wọpọ bi atunwi tabi fo awọn ọrọ nipa gbigbe awọn titete iṣaaju, ati ẹrọ oluyipada autoregressive pẹlu ami iduro ikẹkọ jẹ ki awoṣe mu awọn gbolohun ọrọ gigun-iyipada ni oore-ọfẹ.
Ipa Ilana
Wiwọle ati arọwọto
O ṣe ilọsiwaju iraye si nipasẹ transcription, alaye, ati awọn atọkun ohun.
Iye owo ati isuna
Awọn ẹgbẹ Media le firanṣẹ ohun didan yiyara pẹlu awọn isuna-owo kekere.
Iyara ati iwọn
Awọn ọna ṣiṣe ti nkọju si alabara le ṣe ilana awọn ibaraẹnisọrọ sisọ ni iwọn nla.
Ọjọ iwaju ti Tacotron 2
Apẹrẹ ipele meji ti Tacotron 2 ṣe atilẹyin igbi ti TTS nkankikan. Yiyara ti kii ṣe autoregressive arọpo bi FastSpeech 2 yọkuro decoder lẹsẹsẹ fun iyara ati iduroṣinṣin, ati pe vocoder WaveNet ti wa ni bayi nigbagbogbo paarọ fun HiFi-GAN tabi awọn awoṣe itankale. Aaye naa n lọ si opin-si-opin ni kikun ati agbọrọsọ olona-pupọ, ikosile, ati awọn eto didi ohun-shot, ṣugbọn Tacotron 2 jẹ itọkasi ipilẹ fun awọn opo gigun ti spectrogram.
Real-World imuse
Ngba agbara awọn ohun ti o dun ni agbara ni awọn ọja ati awọn oluranlọwọ ọrọ-si-ọrọ Google
Ṣiṣẹda alaye asọye fun awọn iwe ohun ati adarọ-ese
Pese awọn ohun fun awọn oluka iboju ati sọfitiwia iraye si
Ṣiṣẹ bi ipilẹ iwadi ati apẹẹrẹ ẹkọ fun awọn opo gigun ti TTS ti iṣan
Awọn ewu & Awọn ọna iṣọ
ilokulo ohun ati awọn ewu afarawe ṣe pọ si nigbati igbanilaaye ba sonu.
Yiye le ju silẹ kọja awọn asẹnti, awọn ede-ede, tabi awọn agbegbe alariwo.
Ohun afetigbọ sintetiki le jẹ aṣiṣe fun ọrọ ododo laisi isamisi to yege.
Ilana Ilana imuse
Gba ifọkansi ti o fojuhan fun gbigba ohun, ti ẹda, ati ilotunlo.
Didara idanwo kọja awọn agbohunsoke oniruuru ati awọn ipo abẹlẹ.
Ṣetumo nigbati eniyan gbọdọ ṣe atunyẹwo tabi fọwọsi awọn abajade.
Aami ohun sintetiki ki o tọju awọn igbasilẹ provenance fun iṣiro.
Tesiwaju Ṣiṣawari
Free newsletter
Get the daily AI briefing
Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.
One email each weekday. Unsubscribe in one click. We never sell or share your address.
Test yourself
Take the Tacotron 2 quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
Itọsọna atẹle
Awoṣe Prosody
Awọn ibeere ti a beere nigbagbogbo
Kini Tacotron 2?
Tacotron 2 jẹ opin-si-opin ọrọ-si-ọrọ eto lati Google (2017) ti o yi ọrọ kikọ taara sinu a mel-spectrogram, eyi ti a nkankikan vocoder iyipada sinu igbesi aye ọrọ. O ṣe agbejade awọn gbigbasilẹ ohun ti o ni idije eniyan lori awọn ipilẹ bọtini.
Aṣoju agbedemeji wo ni Tacotron 2 ṣe asọtẹlẹ lati ọrọ?
Tacotron 2's lẹsẹsẹ-si-atẹle nẹtiwọọki ṣe asọtẹlẹ mel-spectrogram kan, eyiti vocoder kan lẹhinna yipada si ohun ohun.
Kini o ṣe iyipada spectrogram Tacotron 2 sinu fọọmu igbi gangan?
Tacotron 2 ṣe afiwe asọtẹlẹ spectrogram rẹ pẹlu vocoder WaveNet ti a yipada lati ṣe agbekalẹ ohun afetigbọ aise ikẹhin.
Iṣoro wo ni ifarabalẹ ipo-iṣe iranlọwọ ṣe idiwọ?
Nipa gbigbe awọn tito tẹlẹ, akiyesi ifarabalẹ ipo dinku awọn ikuna bii awọn ọrọ ti a tun tabi silẹ.
Bawo ni Tacotron 2 ṣe mọ igba lati dawọ ti ipilẹṣẹ ọrọ kan duro?
Awọn autoregressive decoder asọtẹlẹ kan Duro àmi, jẹ ki awọn awoṣe mu awọn gbolohun ọrọ ti o yatọ si ipari.
Iru ara faaji gbogbogbo wo ni apakan ọrọ-si-spectrogram lo?
Tacotron 2 nlo kooduopo-decoder lẹsẹsẹ-si-tẹle awoṣe pẹlu akiyesi awọn kikọ maapu si awọn fireemu spectrogram.