InoteveraGaidhi rinotevera
Parallel WaveGAN Vocoder
AI Yenyaya
Chidzidzo cheAI yeRuzha
HiFi-GAN uye GAN Vocoders
HiFi-GAN inogadzira-adversarial vocoder inoshandura mel-spectrogram kuita yakasvibira odhiyo waveform kanenge ipapo, ichigadzira studio-yemhando yekutaura nekukurumidza kupfuura nguva chaiyo.
Teerera kugwaro iriAI-yakarondedzerwa · ~2 min
Pa peji ino2 min verenga
Pfupiso
Yakave nhanho yekupedzisira yeazvino zvinyorwa-ku-kutaura nekuti inokurumidza, isingaremi, uye yakaoma kusiyanisa kubva kune chaiyo yakarekodhwa.
Kudzika Kwakadzika
Vocoder inhanho yekupedzisira mumapaipi mazhinji eTTS: modhi yakaita seTacotron kana FastSpeech inofanotaura mel-spectrogram (yakabatana mufananidzo wefrequency nekufamba kwenguva), uye vocoder inozadza iwo chaiwo waveform samples. Ekutanga neural vocoder seWaveNet yakanzwika yakakura asi yakagadzira odhiyo sampuli-ne-sample, ichiita kuti inonoke zvinorwadza. HiFi-GAN, yakaburitswa naKong, Kim, naBae muna 2020, yakatsiva iyo autoregressive loop nejenareta rimwe chete rekupa-mberi rakadzidziswa kurwisa. Hunyengeri hwayo hwekushandisa kushandisa rusarura rwakati wandei rwunotonga odhiyo pazvikero zvakasiyana uye pamusoro peakasiyana periodic mapatani, achimanikidza jenareta kuti iwane zvese zviri zviviri manyorerwo akanaka uye pitch periodicity kurudyi. Mhedzisiro yacho ndeye 22 kHz kutaura kwakagadzirwa mazana enguva nekukurumidza pane chaiyo nguva paGPU, ine mhando inokwikwidza pasi-chokwadi odhiyo.
Kuitwa muNyika yeChokwadi
Kugadzira kuburitswa kwekutaura kwevabatsiri uye maapplication ekufambisa ayo anoda mhinduro pasina kunonoka kunonzwika.
Kusimbisa-chaiyo-nguva izwi cloning uye dubbing maturusi uko cloned mel-spectrogram inopihwa mune yakasikwa-inonzwika odhiyo.
Kutyaira mabhuku ekuteerera uye podcast kurondedzera mapuratifomu anogadzira maawa ekutaura nekukurumidza uye zvakachipa.
Kushanda senge waveform nhanho mukati mekuimba-izwi synthesizer uye mimhanzi demos kuburikidza neBigVGAN-maitiro epasirese vocoder.
Ramangwana reHiFi-GAN uye GAN Vocoders
Mavhokodha eGAN anoramba achiwedzera kuita madiki uye nekukurumidza: vazukuru vakaita seBigVGAN vanowedzera anti-aliased activation kuti vawedzere kune vasingaonekwe vaimbi, zviridzwa, nemitauro, nepo UnivNet neVocos zvichisundira kune wepasirese, ese-bhendi synthesis. Kutenderera uye pane-mudziyo akasiyana zvino mhanyisa vocoding mukati memafoni uye mahedhifoni kune yakaderera-latency vabatsiri. Kuwedzera, kupararira uye kuyerera-kufananidza maodhiyo modhi ari kunyungudutswa muGAN-maitiro ega-pass jenareta, achisanganisa kutendeka kwediffusion nekumhanya kweGAN. Tarisira kuti mavhokodha achadzima kuita zvakajairika-chinangwa neural audio codecs inogonesa zvese kutaura nemimhanzi.
Kuzvishandisa
Nei zvichikosha
- Svika uye svika. Inonatsiridza kusvikika kuburikidza nekunyora, kurondedzera, uye mazwi ekubatanidza.
- Mutengo uye bhajeti. Zvikwata zveMedia zvinogona kutumira odhiyo yakakwenenzverwa nekukurumidza nemabhajeti madiki.
- Kumhanya uye chiyero. Masisitimu anotarisana nevatengi anogona kugadzirisa kutaurirana kwekutaura pamwero mukuru.
Chenjera
- Kushandisa izwi zvisizvo uye njodzi dzekuedzesera dzinowedzera kana chibvumirano chisipo.
- Kururama kunogona kudonha mumitauro, mataurirwo, kana nharaunda dzine ruzha.
- Synthetic audio inogona kukanganisa kutaura kwechokwadi isina mavara akajeka.
Matanho anotevera
- Wana mvumo yakajeka yekutora inzwi, kugadzira, uye kushandisa zvakare.
- Yedza mhando pavatauri vakasiyana uye mamiriro ekumashure.
- Tsanangura apo munhu anofanira kuongorora kana kubvumidza zvabuda.
- Label synthetic odhiyo uye chengetedza marekodhi ekuzvidavirira.
Ramba Uchiongorora
Tsamba yemahara
Tora iyo yemazuva ese AI pfupiso
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
Take the HiFi-GAN and GAN Vocoders quiz
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
Mibvunzo inowanzo bvunzwa
Chii chinonzi HiFi-GAN uye GAN Vocoders?
HiFi-GAN inogadzira-adversarial vocoder inoshandura mel-spectrogram kuita yakasvibira odhiyo waveform kanenge ipapo, ichigadzira studio-yemhando yekutaura nekukurumidza kupfuura nguva chaiyo. Yakave nhanho yekupedzisira yeazvino zvinyorwa-ku-kutaura nekuti inokurumidza, isingaremi, uye yakaoma kusiyanisa kubva kune chaiyo yakarekodhwa.
Nderipi basa rekutanga revokoda seHiFi-GAN mune yemavara-kune-kutaura pombi?
Vocoder ndiyo nhanho yekupedzisira iyo inogadzira chaiyo waveform samples kubva kune mel-spectrogram inogadzirwa neacoustic modhi.
Nei HiFi-GAN ichikurumidza zvakanyanya kupfuura yekutanga WaveNet vocoder?
WaveNet yakagadzira odhiyo sampuli-ne-sample (autoregressively), nepo HiFi-GAN's feed-forward jenareta inoburitsa waveform mupfuti imwe, ichiwana nekukurumidza-kupfuura-chaiyo-nguva yekumhanya.
Chii chinobatsira HiFi-GAN's Multi-Period Discriminator inonyatso batsira kubata?
Iyo Multi-Period Discriminator inogadziridza iyo 1D waveform kuita 2D ichishandisa prime-nhamba nguva kuona iyo periodic chimiro chakasungirirwa pakudzika.
Mazwi eGAN anodzidziswa 'zvakasiyana.' Zvinorevei apa?
Mukugadzika kweGAN, jenareta inodzidza kugadzira masaisai echokwadi akakwana kunyengedza masarura erusarura akadzidziswa kutaura chaiwo kubva kudhidhiyo.
Ndeipi yakazotevera vocoder inotambanudza HiFi-GAN kuti iwedzere zvirinani kune asingaonekwe manzwi, kuimba, uye zviridzwa?
BigVGAN inoyera HiFi-GAN uye inowedzera anti-aliased periodic activation, inovandudza generalization kune kunze-kwe-kugovera odhiyo senge kuimba uye zviridzwa.
Ramba uchidzidza
Mazano anoenderana
Mamwe madhairekitori akasarudzirwa nyaya iyi