InoteveraGaidhi rinotevera
Kugadzirira Chinyorwa-ku-Kutaura Dataset
Tekinoroji
Nhungamiro yehunyanzvi
Offline text-to-speech inoshandura mashoko akanyorwa kuita mutauro munharaunda uchishandisa software yakadhaunirodwa uye modhi kana mafaira ezwi.
Teerera kugwaro iriAI-yakarondedzerwa · ~3 min
Piper neKokoro mapurojekiti akasiyana ane magadzirirwo emhando dzakasiyana, kurongedza, kuvharwa kwemitauro, uye zvinodiwa zvehardware, saka enzanisa zvinyorwa zvazvino uye marezinesi eiyo chaiyo nguva yekumhanya uye izwi raunoronga kushandisa.
Mavara-ku-kutaura, kana TTS, mamepu akanyorwa mavara kune yekutaura waveform. Kumhanyisa maitiro pasina Indaneti kunoreva kuti chishandiso chine nguva inodiwa yekumhanyisa uye modhi kana mafaera ezwi pamudziyo wenzvimbo uye haidi kutumira kutaura kwega kwega kune inoitirwa TTS sevhisi. Inovandudza kuwanikwa asi inotamisa kuisirwa, kuchengetedza, uye kuita mabasa kune anoshanda. Piper ipurojekiti yakavhurika-sosi TTS ine manzwi anogona kudhawunirodha uye yemuno inference tooling. Iyo yazvino Piper purojekiti inochengetwa muOHF-Voice repository uye marezenisi ayo uye mapakeji ruzivo zvinofanirwa kutariswa ipapo. Kokoro imhuri yakaparadzana yeTTS ine kadhi remuenzaniso, manzwi, maraibhurari ekufungidzira, uye ruzivo rwemutauro. Mazita haareve zvakafanana modhi yezvivakwa, mhando, kana nharaunda dzinotsigirwa. Shanduro uye kurongedza kuchinja kunogona kuitika, saka shandisa zvinyorwa zvepurojekiti zvazvino kwete chidzidzo chekare. Modhi yezwi haisi chinhu chakafanana neyenguva yekumhanya inoitakura. Mafomati emhando, mafonemizer, maduramazwi, mapaki emitauro, uye zvinoenderana nekuteerera zvese zvinokanganisa kana synthesis inoshanda. Iyo compact modhi inogona kuenderana neyakamanikidzwa CPU mudziyo, nepo imwe ingangoda imwe ndangariro kana kumhanyisa. Kuyera chizvarwa chekumhanya, mutengo wekutanga, ndangariro tsoka, uye odhiyo mhando pane chaiyo yakananga hardware. Quantization kana akasiyana inference backends anogona kuchinja zvese kushanda zvakanaka uye kubuda. Mutauro uye kuvharwa kwezwi kunofanirwa kusimbiswa padanho remuenzaniso. Nguva yekumhanya inotsigira mutauro hazvireve kuti izwi rese rinoutsigira zvakanaka. Mazita akakodzera, zvipfupiso, manhamba, nyora dzemutauro, uye kushandura kodhi zvingada kujairana nemavara kana mitemo yemataurirwo. Miedzo yekuteerera ine ndima dzinomiririra inodzidzisa pane kutonga imwechete demo mutsara. Rezinesi nemvumo zvinoda macheki akasiyana. Kodhi yeprojekiti, uremu hwemuenzaniso, uye data yezwi remunhu inogona kunge iine mazwi akasiyana. Kana uchigadzira izwi remunhu anozivikanwa, wana mvumo uye uteedzere mitemo inoshanda. Offline execution pachayo hairatidze kuti kugovera kana kushandiswa kwekutengesa kunobvumidzwa. Zvinyorwa zvinyorwa, shanduro, uye sarudzo dzezwi usati watumirwa.
A Raspberry Pi inozivisa chimiro che sensor uchishandisa yakadhindwa Piper izwi mushure mekuyera chizvarwa latency uye ndangariro pachishandiso.
Chishandiso chedesktop chinoshandisa modhi yeKokoro munharaunda uye chinotarisa mutauro unotsigirwa uye magadzirirwo ezwi isati yagadzira mitauro yakawanda.
Chikwata chechigadzirwa chinoongorora rezinesi reiyo TTS yekumhanyisa nguva uye inopupurira zvakasiyana zvirevo zvakabatanidzwa kune yega yega izwi rakadhawunirodha kana modhi.
Muverengi anotarisana nemushandisi anotsemura mavara marefu kuita machunks, anochengetedza nyora dzemutauro, uye anotarisa kuenderera kweodhiyo pamiganhu yechunk.
Offline TTS inogona kuramba ichivandudza sezvo tsoka dzemodhi dzichiderera uye nguva dzekumhanya dzichitsigira mamwe marudzi emidziyo. Vashandisi vanogona kuona mutauro wakakura uye sarudzo dzezwi, kunyangwe kuvharika uye mhando icharamba isina kuenzana pamamodheru. Kurongedza kuri nyore kunogona kurerutsa kubatanidza, nepo modhi yakajeka uye metadata yezwi yaizobatsira zvikwata kuongorora mamiriro ekushandiswa. Synthesis yenzvimbo icharamba ichida bvunzo dzekuteerera, kuyerwa kwehardware, kuongororwa kwerezinesi, uye kubata zvine hunyanzvi kwekuzivikanwa kwezwi. Hardware mabhenji anofanirwa kusanganisira echokwadi mavara uye kutonhora kunotanga. Tarisa zvinhu izvi usati washambadzira kutumirwa.
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
Offline text-to-speech inoshandura mashoko akanyorwa kuita mutauro munharaunda uchishandisa software yakadhaunirodwa uye modhi kana mafaira ezwi. Piper neKokoro mapurojekiti akasiyana ane magadzirirwo emhando dzakasiyana, kurongedza, kuvharwa kwemitauro, uye zvinodiwa zvehardware, saka enzanisa zvinyorwa zvazvino uye marezinesi eiyo chaiyo nguva yekumhanya uye izwi raunoronga kushandisa.
Yemunharaunda TTS inogadzira odhiyo waveform kubva kune yakanyorwa zvinyorwa pachishandiso.
Zvinyorwa zvavo uye zvigadzirwa zvinosiyana, saka kugona hakufanirwe kufungidzirwa kuchinjika.
Kodhi, huremu uye data rezwi rinogona kutakura akasiyana rezinesi mazwi.
Real-time factor inokamura synthesis nguva neyakagadzirwa odhiyo nguva; ndangariro uye kutonhora-kutanga kunonoka ndezvimwe zviyero.
Mavara akajairwa uye chunk miganhu anogona kukanganisa echokwadi kubuda.
Ramba uchidzidza
Mamwe madhairekitori akasarudzirwa nyaya iyi
InoteveraGaidhi rinotevera
Kugadzirira Chinyorwa-ku-Kutaura Dataset
Tekinoroji