Umbhalo usuka kumazwi Ama-AI Agents Umculo Ukukhuluma kuMbhalo Ukulungiswa kwezwi Uhlelo lwezwi Imakethe Okuningi
Umbiko wephutha / Umbuzo wezici

Ukufaka umsindo we-AI

Buyisela inqwaba yomsindo ngezwi elihlobene ne-AI-synthesized elifana nezwi elibhekene nalo. Lungisa ukukhishwa okungalungile ngaphandle kokurekhoda kabusha konke.

Ubhalise
Asikho isikhulumi se-TTS ezweni lakho. Sicela usize ukungeza isandla sakho! Uhlu lwamagama

Layisha umsindo kwi-Inpaint

500 amaphawu ngesekondi yesandi esishintshiwe

Thwebula bese ushiya ihele lakho lapha, noma bheka

Supports MP3, WAV, FLAC, OGG, M4A. Max 500 MB (2 GB on paid plans). Up to 10 minutes.

ifayela.mp3

0 MB

Umsuka womsindo — cindezela ukuthola i-take ebi

0.00s / 0.00s

Izilungiselelo zokudweba

0 / 500 amaphawu
Ingabe kude ukuxhuma amaphuzu okuxhuma. 80ms kuyiphutha — ukucushwa komdlalo kubukeka kujwayelekile, akukho ukucindezela okuphindwe kabili okuzwakalayo.
Bhala ngokukhululekileyo ukusebenzisa ukudweba umsindo

Ukuklonya umsindo nokuhlela isishintshi...

Ukuslice → ukuklonya umsindo obhekene → ukuslice nge-crossfade
Uthatha isikhathi eside? Imiphumela yakho izovela ku- generation history lapho ukulungele khona.
Umsindo ophrintiwe ulungile

Phambi (Okungajwayelekile)

Phakathi (Ipeyintiwe)

Layisha phezulu umsindo otholwe

Indlela i-Audio Inpainting isebenza ngayo

Ukudweba kufana nesandi se-Photoshop esizizwayo. Siklone umsindo kusuka kumsindo ojikeleza okukhethiwe, sihlanganise umgwaqo omusha kulo msindo, futhi siyibeke emuva nge-crossfade encane.

Izimpendulo ezinhle kakhulu: shiya imizuzwana engu-3 yokukhuluma uma uhlela indawo ukuze i-cloner ithole okuqukethwe okulungile.

Izincomo zemiphumela engcono kakhulu

  • Gcina umkhawulo ophawulwe ngokuqinile njengoba kungakwenzeka — ukuthatha okubi kuphela
  • Umbhalo oshintshiwe kufanele ube nobude obufanayo nobude obushintshiwe
  • Misela ulwimi ukuze lufane nomsindo womsuka ukuze lufane nomsindo ongcono
  • 80ms crossfade akubonakali; bump to 150ms if you hear a click
  • Ukuhlela okude (>10s), cabanga ngokuphinda urekhode yonke ingxenye

Indlela i-AI Audio Inpainting isebenza ngayo

Ukulungisa okuncane, kuhlanganisa umsindo, ngaphandle kokuphinda urekhode isiqephu.

Isigaba 1

I-Upload + Phawula isigaba

Layisha umsindo wakho bese usebenzisa isici sokususa ukufaka uphawu lokuqamba/kuphelisa isiqephu ofuna ukusishintsha. Bhala umbhalo wokushintsha.

Isigaba 2

Umsindo Clone + Synthesize

Sikhipha kuze kube yimizuzu engu-12 yesandi esicacile esibhekene nokukhethwa kwakho, siklone umsindo womsindo, futhi sihlanganise umgwaqo omusha kulo msindo.

Isigaba 3

Ukuxhuma oku-crossfade

I-clip ehlobene ixhunywe ku-recording yangempela nge-crossfade enamandla afanayo emigwaqweni emibini yokuhlela. Ama-borders angeke azwakale.

Izinhlelo zokudweba umsindo

Qala ngokukhululekile, uthuthukise uma ufuna okuningi

Ikhululekile
  • Kuze kube yimizuzu engu-10 yomthombo wefayela
  • Umbhalo wokushintsha obungaphawu-500
  • 4-sekondi inpaint ngesicelo
  • 80ms crossfade splice
  • I-OpenVoice + CosyVoice 2 ifaka iziphequluli
Okuthandwa kakhulu
I-akhawunti Ekhululekile
  • Kuze kube yimizuzu engu-10 yomthombo wefayela
  • Umbhalo wokushintsha obungama-5,000
  • Ukugqwala okuhlanganisiwe okuhlanganisiwe (0-250ms)
  • Ukushintsha imodeli yomsindo
  • Ukudala imbali + hlela kabusha
Bhala
I-Pro
  • Kuze kube yimizuzu engu-30 yomthombo wefayela
  • Umbhalo wokushintsha obungama-100,000-amaphawu
  • Iphutha le-GPU
  • Ukungena kwe-API (/v1/audio-inpaint/)
  • Ukudweba kweqembu (amabanga ahlukahlukene)
Ukulungiswa

Imibuzo ebuzwa kaningi

Ukudweba umsindo (kwaziwa nangokuthi ukugcwalisa umsindo noma ukudlulisa umsindo) kukuvumela ukuthi ushintshe isigaba se-audio recording esisha ngolimi olusha oluhlanganisiwe nge-AI olufana nomsindo wokuqala. Kuyinto elinganayo nomsindo we-Photoshop's content-aware fill - udwebe phezu kwengxenye ongathandanga, ubhale lokho okudingekayo endaweni, futhi i-AI ikhiqize ushintsho olungenalutho.

Mark the time range to replace, type the new line of dialogue, and click Inpaint. Our AI clones the voice from the audio surrounding your selection, synthesizes the new line in that voice, and splices it back into your recording with a short crossfade so the edit is inaudible.

Sebenzisa uma unegama elilodwa elibi, ukuchaza okungalungile, igama elihamba, igama elikhohlisayo, noma iphutha leqiniso elingenalutho. Ukurekhoda kabusha isigaba esigcwele kwenza ukuthi kungabi khona ukufana kwe-tonal nengxenye encane yephrojekthi — ukudweba kwenza kuphela okudinga ukumiswa ngenkathi ugcina wonke ama-syllaba akhona.

Abasebenzisi abamahhala bangadweba amafayela kuze kube yimizuzu engu-10. Ababhalisile bangadweba amafayela kuze kube yimizuzu engu-30. Umbhalo wokushintshana uphele ngophawu olungu-500 kubasebenzisi abamahhala, 5,000 ku-akhawunti emahhala, no-100,000 ku-akhawunti ezikhokhelwayo.

Kuhle kakhulu. I-AI isebenzisa imizuzwana engu-12 yomsindo ejikeleza ukulungisa njengenkomba yomsindo, okunembile nganoma iyiphi yemodeli yethu ekwazi ukuklonya (OpenVoice, CosyVoice 2) ukuqoqa umsindo womsindo, i-pitch, nesitayela sokukhuluma. Ukuthola imiphumela engcono, shiya imizuzwana engu-3 yomsindo ohlanzekile ngaphambi kokulungisa indawo.

Sisebenzisa i-80ms elinganayo-power crossfade kunoma yiziphi izixhumanisi (ikhanda→ukushintsha kanye nokushintsha→ikhanda) ngokuzenzakalela. Ungayilungisa kusuka ku-0ms (ukucisha okunzima) kuya ku-250ms nge-Crossfade slider. I-crossfades ende ifihla ukuhlela ngokugcwele kodwa ingaxhumana ngokuzwakalayo nama-word alinganayo emkhawulweni.

Ukudweba umsindo kulandela ukuhlanganisa ulwimi olufanayo njengenhlamvu yezwi. Sikhetha ngokuzenzakalela i-OpenVoice ngezinye izilimi futhi i-CosyVoice 2 ngesi-Chinese, isi-Japanese, nesi-Korean. Ungashintsha imodeli emininingwaneni ephakeme.

Ukhokhiswa ama-characters angu-500 ngomzuzwana ngamunye wesandi esishintshiwe. Ukulungisa okungu-4-umzuzwana kubiza ama-characters angu-2,000. Izindleko ziyimfihlo ngosayizi wokuba umbhalo oshintshiwe ude kangakanani, njengoba ukuxubha okungaphansi kwe-clone kulawulwa isikhathi sokusebenza se-clip entsha, hhayi ubude bombhalo.

Ngokuya ngemigomo yethu yenkonzo, ungadweba kuphela umsindo owunikazi wakho noma onesivume esicacile sokuhlela. Ukuletha izixhumanisi ezingezinhle, okuqukethwe okukhohlisayo, noma ukudalula isithombe kuvunyelwe. Sibeka i-watermark kumsindo owenziwe futhi sibhale zonke imisebenzi yokudweba ukuze sibuyekeze ukubhekwa kokusetshenziswa kabi.

Ukucisha i-clip kwenza kube nethuba eliphawulekayo lokuhamba nokuphefumula; ukufakelwa okufanayo kwezintathu kwenza kube nethuba lokungalingani kwe-tonal. Ukudweba kwenza kube nethuba lokukhuluma elifana nezwi elibhekene nalo, ngakho-ke abalalela balalela umsindo oqhubekayo, obukekayo.

Yebo — POST ku /v1/audio-inpaint/ ngefayela lomsindo, start_sec, end_sec, kanye ne replace_text. Ingxenye ephelile ibuyisela umsebenzi UUID; umbuzo /v1/speech/results/?uuid= ukubuyisela umsindo otholwe uma ulungile. Bona i-API docs ngezinkethe.

I-ElevenLabs Speech-to-Speech ivuselela yonke i-line yomsindo kusuka ekuqaleni kohlelo lwezwi elifunayo. Ukudweba kwesandi sikwenza ngokuzivocavoca: ihlela kuphela umkhawulo ophawulwe, igcina wonke ama-bytes okugcina okungenakuchofozwa, futhi ilinganise i-clip entsha nomsindo obhekene nayo ngaphezu kwe-library yomsindo ehlukile.
5.0/5 (1)

Yini esingayithuthukisa? Umbono wakho usiza ukuxazulula izinkinga.

Misa umsindo wakho emaminithini

Buyisela noma iyiphi ingxenye yokufaka umsindo ngezwi elihlobene ne-AI elifana nomsindo oyinhloko. Bhala ubhalise ukuze uqale.

Ubhalise Bonisa ukuthengwa