InoteveraGaidhi rinotevera
FastPitch Pitch-inodzora TTS
AI Yenyaya
Chidzidzo cheAI yeRuzha
Glow-TTS imhando yemavara-kune-kutaura modhi inodzidza kuwiriranisa mavara kune kutaura pachayo uchishandisa yakangwara yekutsvaga hungwaru, ichibvisa kudiwa kweanorongeka.
Teerera kugwaro iriAI-yakarondedzerwa · ~2 min
Izvo zvine basa nekuti zvinoita kuti kudzidzisa kuve nyore uye synthesis nekukurumidza uye kuenderana.
Glow-TTS, yakaunzwa naKim uye vaanoshanda navo muna 2020, inogadzira mel-spectrogram kubva kumavara vachishandisa inoyerera-yakavakirwa decoder uye yakavakirwa-mukati yekumisikidza nzira inonzi Monotonic Alignment Search (MAS). Pakutanga maTTS masisitimu akaita seTacotron 2 akashandisa kutarisa kuti asarudze kuti ndeupi mavara anoenderana neodhiyo furemu, asi kutarisa kunogona kusvetuka mazwi, kuadzokorora, kana kutyora mitsara mirefu. Glow-TTS pachinzvimbo inotora kurongeka kunofanirwa kuve monotonic (chinyorwa chinoverengwa kuruboshwe-kuenda-kurudyi) uye kufungidzira (ese mamepu echiratidzo chemepu kune imwechete furemu). Inoshandisa hurongwa hwekuita kuti uwane iyo ingangoita kurongeka kwakadaro panguva yekudzidziswa, ipapo diki yenguva yekufungidzira inodzidza kuigadzira pakufungidzira. Izvi zvinopa kugadzirwa kwekutaura kwakasimba, kwakafanana, uye kunodzoreka.
Kudzidzira rakasimba redhiyo bhuku remutauri izwi risinga svetuka kana kudzokorora mazwi pandima refu
Kusimbisa nhanho yekuenzanisa yeVITS-yakavakirwa-yakavhurika-sosi izwi vabatsiri uye vanoverenga skrini
Kuvaka inodzoreka TTS kwaunotambanudza kana kudzvanya nguva dzefonime kune inononoka, mataurirwo akajeka mumapurogiramu ekudzidza mutauro.
Kugadzira dhatabheti dzematauriro emitauro ine zviwanikwa zvishoma apo data inobatanidzwa nemaoko iri shoma
Iyo monotonic alignment zano rakapayona neGlow-TTS ikozvino rinotsigira akawanda emazuva ano asiri-autoregressive masisitimu, anosanganisira VITS, ayo anoisanganisa nevocoder yekuguma-kusvika-kumagumo waveform chizvarwa. Tarisira kuenderera mberi kwekushandiswa kweMAS-style yakaoma kurongeka mumitauro-yepasi-zvishandiso, chaiyo-nguva-pa-mudziyo manzwi, uye inodzoreka kutaura uko nguva, kukwirira, uye kufamba kunofanirwa kupepetwa zvakajeka. Diffusion uye kuyerera-kufananidza TTS inowedzera kukwereta iyi yakachena mavara-kune-frame mepu kuti igadzikane.
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
Glow-TTS imhando yemavara-kune-kutaura modhi inodzidza kuwiriranisa mavara kune kutaura pachayo uchishandisa yakangwara yekutsvaga hungwaru, ichibvisa kudiwa kweanorongeka. Izvo zvine basa nekuti zvinoita kuti kudzidzisa kuve nyore uye synthesis nekukurumidza uye kuenderana.
Tarisiro inogona kukanganisa mapinji akareba, zvichikonzera kusvetuka kana kudzokororwa mazwi; Glow-TTS inomanikidza monotonic kurongeka kudzivirira izvi.
MAS ndeye Monotonic Alignment Search, iyo ine simba-programming routing inowana yakanakisa mavara-kune-frame kurongeka.
Hurukuro inoverenga zvinyorwa zvakatevedzana, saka kurongeka kunofanirwa kuenderera mberi nemukati sezvo nguva ichienda.
Kupenya-TTS inoshandisa inoyerera-yakavakirwa decoder, iyo inopa chaiyo mukana wekuti MAS inogona kukwirisa pamusoro pekumisikidza.
MAS inongodiwa mukudzidziswa; dhizaini yenguva yakadzidzwa inopa per-token furemu inoverengera pakufungidzira, ichigonesa chizvarwa chakafanana.
Ramba uchidzidza
Mamwe madhairekitori akasarudzirwa nyaya iyi
InoteveraGaidhi rinotevera
FastPitch Pitch-inodzora TTS
AI Yenyaya