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

Isici sokukhipha umsindo we-AI

Yakha ama-tracks karaoke, ama-acapellas, nama-tracks wokuxhasa mahhala.

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

Layisha phezulu umculo kwinqubo

3,000 amaphawu ngalinye

Thwebula bese ushiya ihele lakho lapha, noma bheka

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

ifayela.mp3

0 MB
— noma urekhode kusuka kumsindo wakho —
00:00

Izicwangciso zokuhlukanisa

Izimo ezimbili zisebenzisa imodeli efanayo yeMel-Band RoFormer.
Imahhala ukuzama nge-akhawunti emahhala

Indlela isebenza ngayo

Layisha phezulu umculo wakho
I-AI ihlukanisa amabhayisikobho
Layisha phezulu

Ukuhlukanisa ama-vocals kusuka kuma-instrumentals...

Sebenzisa Mel-Band RoFormer ku Ilinganiswe quality
Uthatha isikhathi eside? Imiphumela yakho izovela ku- generation history lapho ukulungele khona.
Ukuhlukaniswa kwezwi kuqediwe

Ama-vocals

0:00

I-Instrumental

0:00
Uhlobo lokuhlukanisa 8.2 dB SDR
Ukushaya kwezwi < 2%
Isikhathi sokudlalwa kwengoma 3:42
Isikhathi sokuphatha 12.4s

Isici sokuzihlukanisa

Imodeli Mel-Band RoFormer
I-Output Izwi + i-instrumental
I-Vocal SDR 11.3 dB
I-SDR yezinhlamvu 16.0 dB

Ilinganiselwe ku-50 MUSDB18 ukuhlolwa okuncane. SDR yisignal-to-distortion ratio: phezulu kuhlanzekile. Injini yethu yangaphambilini ithole 8.4 dB (izwi) ne-13.1 dB (izinhlamvu).

Kungani TTS.ai vs Abancintisana

Izinga elikhululekile likhona

Zama nge-akhawunti emahhala. Abancintisana njenge-LALAL.AI, Moises, ne-PhonicMind bakhokhisa i- $ 5-25 / ngenyanga yokususa umsindo oyinhloko.

Uhlobo lokuhlukaniswa okulinganiselwe

Izwi lihlukaniswa ngemodeli yeMel-Band RoFormer. Ku-50 MUSDB18 ukuhlolwa okuqoshwe, ithole i-11.3 dB yezwi SDR ne-16.0 dB ye-instrumental SDR, cishe i-3 dB ngaphezulu kwenjini yethu yangaphambilini kuzo zombili.

Akunamanzi noma ama-caps wekhwalithi

Layisha phezulu okuqukethwe okusezingeni eliphelele ku-WAV, MP3, noma FLAC. Akunamanzi, akukho ukucisha, akukho ukuvimbela i-bitrate kunoma iyiphi i-tier.

Ukungena kwe-API ukuphatha iqembu

Inqubo amakholo ama-albhamu ngokuzenzakalela nge-REST API yethu. Ilungile kuma-company akhiqiza umculo kanye nama-DJ pools.

Izincomo zemiphumela engcono kakhulu

  • Sebenzisa umngenisi we-WAV noma we-FLAC ukwehlukanisa umgangatho ophezulu
  • Khetha zombili ukuze uthole i-acapella ne-instrumental kusuka ku-upload eyodwa
  • Ama-studio masters ahlukanise ngokucophelela kunalokho okurekhodiwe noma okubhuthayo
  • Amaculo ane-stereo ehlukaniswe ngokucacile akhiqiza izimpendulo ezinhle kakhulu
  • I-MP3s ecindezelwe kakhulu (128kbps noma ngaphantsi) ingase ibe nekhwalithi yokuhlukaniswa ephansi
  • Ukuhlukanisa izihlahla eziningi (ama-drums, ama-bass, njll.), sebenzisa i- Isihlukanisi somsindo

Indlela isebenza ngayo

Inethiwekhi enzulu ye-neural ibheka i-audio spectrum yengoma yakho, ikhomba ukusakazwa kwezwi, futhi ihlukanise ngokucophelela kusuka ku-instrumental backing ngezindlela ezintathu.

Isigaba 1

Layisha phezulu umculo wakho

Dlulisa bese ushiya noma iyiphi ingoma ku MP3, WAV, FLAC, OGG, noma M4A. Sixhasa amafayela afinyelela ku-50MB nama-albhamu afinyelela kumizuzu engu-10. Ifayela lakho libhalwe ngokufihliwe ngesikhathi sokulayisha futhi liqhubekelwa phambili ku-GPU ehlukanisiwe. Amafayela asuswa ngokuzenzakalela ngaphakathi kwehora elilodwa lokuqhubekela phambili.

Isigaba 2

I-AI ihlukanisa amabhayisikobho

Imodeli yeMel-Band RoFormer iguqula umsindo wakho ube yi-spectrogram ehlukaniswe ibe yi-mel-scaled frequency bands, bese isebenzisa ukunakekelwa kwe-transformer ngesikhathi nesikhathi sokulinganisela umsindo. I-instrumental iyimixholo yokuqala ne-vocal track esuswa, ngakho-ke ama-tracks amabili angeza emuva ku-song yakho.

Isigaba 3

Layisha phezulu

Lindela ama-vocal ahlukanisiwe nama-instrumental tracks ahlukanisiwe ngakwesokudla nge-audio players ethu afakwe ngaphakathi, bese ulanda wonke ama-tracks ngefomethi oyifunayo.

Isibonelo sokusetshenziswa kwe-vocal remover

Ukususa umsindo we-AI kuvula izimo ezisha zabalingisi, abaculi, abakhiqizi bezinto eziqukethwe, nabangani bomculo ezingeni elikhulu lezinhlelo zokusebenza.

I-Karaoke

Dala ama-karaoke amasha nganoma iyiphi ingoma ngokushesha. Susa ama-vocals aphambili ukudansa kanye nokuzivocavoca okujwayelekile. Kulungile ngemini karaoke, ukuqeqeshwa kokuzivocavoca, ukuphumula kwezwi, ukuqeqeshwa kwekhoriya, nama-karaoke bars. Isebenza nganoma iyiphi inhlanganisela kusuka ku-pop kuya ku-rock kuya ku-R&B.

Ukuxhuma kabusha & Mashups

Thola izingoma ezihlukanisiwe noma iziculo ezisetshenziswayo ukwenza imisindo emisha, ama-mashups, kanye nokuhlela i-bootleg. Yenza imisindo emisha kusuka kungoma eyodwa ne-beat kusuka kwenye. Faka ama-acapellas phezu kokukhiqizwa kwakho. Kubalulekile kubakhiqizi abakwenza imisindo esemthethweni nengasemthethweni.

Ukukhishwa komculo

Sebenzisa izimbobo zezwi kanye nezingxenye zezinhlamvu ezikhona ezikhona ukuze zisetshenziswa emikhiqizweni entsha. Thola izimbobo zezwi, ama-ad-libs, nama-harmonies. Khubaza ama-gitar riffs, ama-bass lines, noma ama-synth patches ukushayela ukushayela. Yikhawulele umfutho wakho wokusebenza wemboni ngokujulile.

I-DJ Sets & Live Performance

Dala amathuluzi ahlanzekile nama-acapellas alungele amasethingi we-DJ aphilayo. Dala ama-transition tracks, yenza ama-mashup-ready stems, futhi ulungiselele ama-vocal drops alungele ukusakazwa okuphilayo. Kubalulekile ku-DJs abafuna ama-mixtures ahlukile amisa amasethingi abo ngokuhlukile ku-competition.

Ukufundisa umculo

Ukwenza iziqephu zezwi ezihlukile zokuzivocavoca, ukuqeqeshwa kwendlebe, nokufundela isayensi yomculo. Ukwenza uharmonizing ngemiculo ehlukanisiwe. Ukuhlola izindlela zezwi ngokulalela imiculo ehlukanisiwe. Ukwenza imiculo ehlukanisiwe yabafundi bemiculo kanye nabaqeqeshi bezwi.

Ukwakha ividiyo nesihloko

Thola amathuluzi omculo wendawo yangasese kumavidiyo, amapodcasts, kanye ne-social media content. Sebenzisa iziguqulo ezingena-vocal zengoma ezithandwayo njengendawo yangasese ye-tracks ngaphandle kokukhathazeka kwe-lyrical. Dala izimbobo ze-soundbeds ezikhethekile ze-YouTube, TikTok, ne-Instagram content.

Izingoma zesigqoko

Rekoda izingoma zakho ezingenalutho phezu kwe-instrumental track yangempela yesiphi na isingoma. Dala izibuyekezo zesigqoko esisezingeni eliphakeme nge-backtracks ethembekile. Ilungile kubaculi abathuthukisa i-portfolio yabo, ukurekhoda i-demo, noma ukushicilela izigqoko ku-YouTube nemidiya yomphakathi.

Ukudluliswa & Ukuhlelwa

Ihlukanise ama-vocals ukuze ubhale ngokucophelela i-lyric, ikakhulukazi ama-albhamu anezinhlamvu ezinzima ezikwenza kube nzima ukulalela amagama. Khubaza ama-melodies, ama-harmonies, nama-ad-libs ngokucophelela. Kusetshenziswa abaculi nababhali bama-musicologists abafunda ukudweba kwezwi.

I-Best Free Vocal Remover Online

Isebenza nge Mel-Band RoFormer

Ukususwa kwezwi kuqhuba imodeli ye Mel-Band RoFormer, i-transformer ehamba phakathi kwesikhathi kanye ne-mel-scaled frequency bands ye-spectrogram. Ku-50 MUSDB18 ukuhlolwa okuqoshwe kulinganiselwe ku-11.3 dB vocal SDR kanye ne-16.0 dB instrumental SDR (uphawu-ku-ukuphambukisa ukulingana, okuphezulu kuhlanzekile), cishe ku-3 dB okungcono kuzo zombili kunenjini esiyisebenzisa ngaphambili. Kulindeleke ukuthi kube ne-bleed ezindaweni eziqinile noma ezicindezelwe kakhulu.

Injini eyodwa kunoma yisiphi isikhala

Uhlukaniso lwe-akhawunti ne-free lusebenza ngemodeli efanayo, ngakho-ke imiphumela ye-free ayiyona imboniso esezingeni eliphansi. Indlela okhetha ngayo ishintsha indlela umsebenzi okhokhwa ngayo, hhayi ukuthi imodeli isebenza kanjani.

Akukho phawu lwamanzi, Akukho mkhawulo

Ngokungafani nezinye izithuluzi eziningi zokususa umsindo ezingeza ama-watermarks, zivimbela umgangatho we-output ezinyangeni ezikhululekile, noma zivimbela amafomethi wokulayisha, i-TTS.ai ikunikeza amafayela wokukhipha ahlanzekile, angabanjwanga. Layisha phezulu ku-MP3, WAV, noma FLAC ngekhwalithi egcwele. Amabhayisikobho akho ahlukanisiwe alungele ukusebenzisa ngokushesha kunoma iyiphi i-DAW noma umdlali wemidiya ngaphandle kokusebenzisa noma yikuphi ukwenziwa kwe-post-processing okudingekayo.

I-REST API yabathuthukisi

Ukufaka ukususwa kwezwi kuzinhlelo zakho zokusebenza nge-REST API yethu elungele abathuthukisi. Ukwenza ama-albhamu ngokuzenzakalela, ukwakha izici zokususa izwi kuma-apps omculo, ukwakha ukujikeleza kwemisebenzi yokulungiselela i-DJ, noma ukwakha amathuluzi ezinhlelo zokuqeqesha umculo. Ukungena kwe-API kufakwe kunoma iyiphi i-plan. SDKs ye-Python, i-JavaScript, kanye nama-cURL amasibonelo ku-documentation.

Ukususa izinhlelo zezwi

Qala ngokukhululekile, uthuthukise uma ufuna okuningi

Ikhululekile
  • Former vocal model
  • Amangoma angama-10-imizuzu
  • Izingxoxo + Izinhlamvu
  • I-MP3 & WAV output
  • Akukho akhawunti edingekayo
Okuthandwa kakhulu
I-akhawunti Ekhululekile
  • Imodeli efanayo ne-free level
  • 4-stem ihlukaniswa ku-Stem Splitter
  • Ilula / Ilinganiswe / Imodi engcono kakhulu
  • Zonke izinhlelo zesiphumezi
  • 15,000 amaphawu amahhala
Bhala
I-Pro
  • Amafayela omsindo ade kakhulu
  • Ukuhlukaniswa kweqembu
  • Ukufinyelela kwe-API
  • Ukuphathwa kwesinqumo
Ukulungiswa

Imibuzo ebuzwa kaningi

Ukukhipha umsindo we-AI sebenzisa imodeli yokufunda ejulile (Mel-Band RoFormer) ebheka i-spectrogram yengoma yakho ngesikhathi kanye ne-mel-scaled frequency bands. Ilinganisela umsindo we-voice track, futhi i-instrumental iyi-original mix with that vocal track removed, producing clean karaoke-style output.

Imodeli yokukhuluma ilinganiselwe 11.3 dB yokukhuluma SDR ne 16.0 dB yezinhlamvu SDR ku-50 MUSDB18 ukuhlolwa okuncane, cishe 3 dB okungcono kuzo zombili kunenjini esiyisebenzisayo ngaphambili. I-instrumental output ivame ukuba mhlophe ngokwanele ku-karaoke noma ukuxuba. Ezinye izimo zokugqwala zingase zivele eziqenjini ezicutshungulwe kakhulu noma ezicindezelwe.

Yebo! Uthola wonke ama-vocals ahlukanisiwe kanye ne-instrumental track. Le nto isetshenziswa ukwenza ama-acapella versions, ukuqoqa ama-vocals ama-remix, noma ukudlulisa ama-lyrics kusuka kuma-mixes anzima.

Sixhasa amafayela we-MP3, WAV, FLAC, OGG, M4A, ne-WEBM angaba ngu-50MB. Ukuthola imiphumela engcono, sebenzisa ifayela lomsuka elisezingeni eliphakeme elitholakalayo (WAV noma FLAC lithandeka kune-MP3 ecindezelwe).

Yebo, ukwenza ama-tracks karaoke kuyinto eyodwa ethandwa kakhulu. I-instrumental output ihlanzekile ngokwanele ukukhombisa karaoke, ukulalela amavidiyo, kanye nomculo wesizinda sezinto eziphilayo.

Layisha phezulu umculo futhi ithuluzi lizokhiqiza ngokuzenzakalela i-vocal track ehlukanisiwe (acapella) kanye ne-instrumental. Ukukhishwa kwe-vocal kusebenza kahle kakhulu kuma-tracks axhunywe kahle nama-vocals acacile. Sebenzisa amafayela womthombo we-WAV noma we-FLAC ukuze uthole izimpendulo ezicacile.

Uhlelo lokuphatha i-batch lutholakala nge-API yethu, okuvumela ukuthi ufake amabhayisikobho amaningi ukususwa kwezwi kuhlelo lokusebenzela oluzenzakalelayo. Uhlelo lokuphathwa kwe-web lusebenza ngomculo owodwa ngasikhathi sinye ngemboniso osheshayo nokukhuphela.

Ithuluzi lethu lisebenzisa imodeli yeMel-Band RoFormer yokuhlukanisa umsindo. Ku-50 MUSDB18 ukuhlolwa okuncane, ilinganisela i-11.3 dB ye-SDR yomsindo kanye ne-16.0 dB ye-SDR ye-instrumental. Ukuhlukanisa okusekelwe ku-transformer kwenza kube ne-artifacts encane kunaleyo eyenziwe ngezindlela ezidala ezisekelwe ku-spectral.

Yebo, kodwa izimpendulo zihluka ngokulalela okuqhubekayo. Ukurekhoda kwestudio kunikeza ubuhle obungcono bokuqedwa. Ukurekhoda okuqhubekayo nge-noise yomsakazo, i-reverb, kanye nezisetshenziswa ezidlulisayo kungase kube ne-bleed-through eningi, kepha imodeli ikhiqiza izimpendulo ezisebenzisekayo.

Ukususa umsindo kusetshenziswa amaphawu angu-2,000 ngalinye lethonya eliphathwayo. Ama-akhawunti amahhala athola amaphawu angu-15,000 uma ebhalisa. Ithuluzi litholakala kuwo wonke ama-plans akhokhelwayo ngezinto ezikhokhelwayo ezigcwele zokusetshenziswa okujwayelekile.

Umsindo ohlukaniswe nganoma iyiphi i-plan ekhokhelwa ungasetshenziswa ngokuhweba, futhi ukuthuthukiswa kufaka umsindo ohlukaniswe ngaphambili ku-free level; i-free-tier output isetshenziswa kuphela umuntu, hhayi ukuhweba. I Mel-Band RoFormer vocal model ikhishwa ngaphansi kwelayisense ye-MIT, ngakho-ke imodeli ifaka akukho kunqanyulwa ekusebenziseni okuhweba. Kuzo zonke izimo kufanele ube nelungelo lomdlalo oyinhloko: ukusebenzisa umculo osemthethweni ngaphandle kwegunya kungalimaza umthetho we-copyright kungakhathaliseki ukuthi umsindo usebenza kanjani.

Isikhathi sokucubungula sixhomekeke kude kwe-track. Ingoma ejwayelekile yemizuzu engu-3-5 ithatha imizuzwana engu-15-45 ukucubungula. Ama-tracks ade kakhulu noma ama-mode wokuhlukanisa osezingeni eliphakeme angathatha imizuzu embalwa. Iziphetho zidlalwa ngokushesha kwi-browser yakho.
5.0/5 (1)

Yini esingayithuthukisa? Umbono wakho usiza ukuxazulula izinkinga.

Susa ama-vocals kusuka kunoma iyiphi indima ngaphandle kwezindleko

Join thousands of musicians and creators using TTS.ai. Vocal removal with a Mel-Band RoFormer model on every tier. 15,000 free characters with signup.

Ubhalise Bonisa ukuthengwa