InoteveraGaidhi rinotevera
Prosody Modelling
AI Yenyaya
Chidzidzo cheAI yeRuzha
Tacotron 2 is an end-to-end text-to-speech system kubva Google (2017) inoshandura zvinyorwa zvakanyorwa zvakananga kuita mel-spectrogram, iyo neural vocoder inoshandura kuita kutaura kwehupenyu.
Teerera kugwaro iriAI-yakarondedzerwa · ~2 min
It produced audio rivaling human recordings on key benchmarks.
Tacotron 2 ine zvikamu zviviri zvikuru. Chekutanga, kutevedzana-kune-kutevedzana network ine hanya inoverenga mavara ezvinyorwa uye inofanotaura mel-spectrogram furemu nefuremu. Encoder inosandura mavara kuita maratidziro akavanzika, chigadziriso chinotarisisa nzvimbo chinoyananisa mavara nemafuremu ekuteerera, uye autoregressive decoder inoburitsa spectrogram apo 'stop token' inodzidza panopera kutaura. Chechipiri, yakagadziridzwa WaveNet vocoder inoshandura iyo mel-spectrogram kuita mbishi waveform. Nekuparadzanisa dambudziko nenzira iyi, Tacotron 2 inodzidza prosody, mataurirwo, uye kufamba kubva kudhata ine kushoma-injiniya ruoko. Iyo yakawana chirevo chepfungwa chibodzwa padhuze nehunyanzvi kurekodha, zvichiita kuti ive chiratidzo mune zvakasikwa-inonzwika synthesis uye template yeyakazotevera neural TTS.
Kusimbisa manzwi anonzwika mu Google's text-to-kutaura zvigadzirwa nevabatsiri
Kugadzira rondedzero inonakidza yemabhuku ekuteerera uye podcasts
Kupa manzwi evaverengi vescreen uye software yekuwanika
Kushanda seyekutanga yekutsvagisa uye muenzaniso wekudzidzisa kune neural TTS mapaipi
Tacotron 2's maviri-nhanho dhizaini yakafuridzira wave neural TTS. Nekukurumidza vasiri-autoregressive vatevedzeri vakaita seFastSpeech 2 yakabvisa inoteedzana decoder yekumhanyisa uye kugadzikana, uye iyo WaveNet vocoder ikozvino inowanzochinjirwa kuHiFi-GAN kana mamodheru. Munda uri kufamba wakananga kumagumo-kusvika-kumagumo uye akawanda-mutauri, anobuditsa, uye zero-kupfura izwi cloning masisitimu, asi Tacotron 2 inoramba iri nheyo yereferensi yemapombi e spectrogram.
Tsamba yemahara
Nhatu dzakasimbiswa dzeAI nyaya mangwanani ega ega pakati pevhiki, dzakanyorwa neChirungu chakajeka. Yemahara zvachose, hapana ads.
Imwe email zuva rega rega revhiki. Bvisa kunyoresa kamwe chete. Hatimbotengese kana kugovera kero yako.
Zviedze iwe pachako
Mhinduro yepakarepo pamhinduro yega yega, uye chitupa chinogovaniswa chine ID inobatika kana wapedza kosi.
Tsigira dzidzo yemahara yeAI. AI Understanding ndeye 501(c)(3) isingabatsiri - hapana ads, hapana paywall, zvachose. Ita chipo
Tacotron 2 is an end-to-end text-to-speech system kubva Google (2017) inoshandura zvinyorwa zvakanyorwa zvakananga kuita mel-spectrogram, iyo neural vocoder inoshandura kuita kutaura kwehupenyu. Yakaburitsa odhiyo inokwikwidza rekodhi revanhu pane akakosha mabhenji.
Tacotron 2's sequence-to-sequence network inofanotaura mel-spectrogram, iyo vocoder yobva yachinja kuita odhiyo.
Tacotron 2 pairi yayo spectrogram predictor ine yakagadziridzwa WaveNet vocoder kuti ibudise yekupedzisira yakaomeswa odhiyo.
Nekutarisisa marongero ekare, kutarisa-kwenzvimbo kunoderedza kukundikana senge mazwi ari kudzokororwa kana kudonhedzwa.
Tacotron 2 predicts a stop token alongside mel frames; generation ends when the learned stopping condition is met.
Tacotron 2 inoshandisa encoder-decoder sequence-to-sequence modhi ine tarisiro kumamepu mavara kune spectrogram mafuremu.
Ramba uchidzidza
Mamwe madhairekitori akasarudzirwa nyaya iyi
InoteveraGaidhi rinotevera
Prosody Modelling
AI Yenyaya