Qwen-Audio-3.1 Uyasebenza

I-Qwen-Audio-3.1 iwuchungechunge lwamamodeli amakhulu ezwi olwakhiwe nguTongyi Qianwen, oluqukethe amamodeli amahlanu: i-ASR yokuqaphela inkulumo, i-ASR-Next yokuqonda umsindo, i-TTS yokwenziwa kwenkulumo, i-TTS-Next yokudala umsindo, ne-Realtime yokusebenzisana ngesikhathi sangempela, oluhlanganisa lonke uchungechunge lokuqonda, ukukhiqiza, ukusebenzisana nokudala. I-Qwen-Audio-3.1 isekela izilimi ezingu-30 nezilimi zesigodi zesiShayina ezingu-16, ukubhala ngokwehlukanisa izikhulumi, ukuqaphela imizwa nengxoxo yesikhathi sangempela enezindlela ezimbili, futhi ingabiza amathuluzi ukuze iqedele imisebenzi.

I-ASR yokuqaphela inkulumo: Isekela ukubhala ngokwehlukanisa izikhulumi ezilimini ezingu-30 nasezilimini zesigodi zesiShayina ezingu-16, inamathegi ezikhulumi nezitempustampu, futhi ingalungisa umbhalo ngokwemvelo.
I-ASR-Next yokuqonda umsindo: Isuka embhalweni odluliselwe ekubhalweni iye ekuqondeni imizwa yabantu, imisindo yemvelo nemisindo yemishini, futhi isekela incazelo yomsindo, ukuthola indawo yomcimbi nemibuzo nezimpendulo zomsindo.
I-TTS yokwenziwa kwenkulumo: Isekela ukwenziwa kwezilimi eziningi nezilimi zesigodi eziningi, idlulise ithoni yezwi ngokwemvelo phakathi kwezilimi, futhi ilawule imizwa nesivinini sokukhuluma ngemiyalelo.
I-TTS-Next yokudala umsindo: Ikhiqiza kanye umsindo ophelele ohlanganisa izwi lomuntu, imisindo ekhethekile nemisindo yemvelo, ofanele ama-podcast, izincwadi ezilalelwayo, amafilimu, imidlalo nezikhangiso.
I-Realtime yokusebenzisana ngesikhathi sangempela: Ilalela futhi ikhulume ngesikhathi esisodwa ngezindlela ezimbili, iqonde imizwa nezinhloso, ishintshe phakathi kwezilimi eziningi, futhi ibize amathuluzi ukuze iqedele imisebenzi phakathi nengxoxo.

Ukumodela okuhlangene kokuphela kuya ekupheleni (ASR): I-Qwen-Audio-3.1-ASR isebenzisa indlela yokuphela kuya ekupheleni, ikhiphe ndawonye amathegi ezikhulumi, izitempustampu nombhalo odluliselwe ekubhalweni, esikhundleni sokucubungula ngezigaba kohlelo oluvamile oluxhumanisiwe; ngenxa yalokho ingagcina iziphazamiso ezimfushane nezinkulumo ezigqagqene. Imodeli iphinde ibe namandla okususa amagama okugcwalisa, ukuphindaphinda, ukuzilungisa nokuhlela kabusha incazelo, okwenza umphumela wokubhala ulungele ngokuqondile amarekhodi emihlangano nokuhlaziywa kwezingxoxo.
Isakhiwo sesizukulwane esilandelayo sokuqonda umsindo (ASR-Next): I-ASR-Next isekelwe esakhiweni sesizukulwane esilandelayo, yandisa into ecubungulwayo isuke “embhalweni osenkulumweni” iye esignali yomsindo ephelele. Ngokuqonda imizwa esezingeni le-acoustic, imisindo yemvelo nemisindo yemishini, iqeda ukuthola, ukuchaza nokucabanga ngemicimbi yomsindo, okubonisa ukuthi i-ASR isuka ethuluzini lokubhala iye kumodeli yokuqonda umsindo ojwayelekile.
Ukudluliswa kwethoni yezwi nokulawula indlela yokusho (TTS): Imodeli ye-TTS yeqa indlela evamile egxile kuphela ekunembeni kokuphinyiselwa kwamagama. Ikwazi ukudlulisa ithoni yezwi ngokwemvelo phakathi kwezilimi nezilimi zesigodi, iphinde imodele imizwa, isivinini sokukhuluma nendlela yokusho ngemiyalelo, ukuze inkulumo eyenziwe ihambisane nokuqukethwe nesimo.
Ukukhiqizwa komsindo okuhlanganisiwe (TTS-Next): I-TTS-Next isebenzisa izinhlobo ezintathu zokufaka, okuwumbhalo, izitempustampu nomsindo oyisithenjwa, ukuze ikhiqize ndawonye izwi lomuntu, imisindo ekhethekile nemisindo yemvelo. Ezingxoxweni ezinemijikelezo eminingi igcina ithoni yezwi nemizwa yabalingiswa ihambisana, iphinde ilawule inqubo yokukhiqiza ngolimi lwemvelo. Isekela ukulawula okunemininingwane kwezitempustampu, ukukopisha izwi nokukhishwa kwekhwalithi ephezulu engu-48kHz.
Isakhiwo esinezindlela ezimbili esigcwele esifundiswe othisha abaningi (Realtime): I-Realtime isebenzisa isakhiwo sokufundiswa othisha abaningi, ithuthukise amandla okuqonda, ukucabanga, ukuveza nokuphepha ngesikhathi igcina ukubambezeleka kuphansi. Lokhu kufezekisa ukusebenzisana ngezindlela ezimbili, lapho ukukhuluma nokulalela kwenzeka ngesikhathi esisodwa, futhi kusekela ukuphazanyiswa noma nini. Imodeli ihlanganisa amandla okubiza amathuluzi, ikwazi ukuqonda izidingo ezishintshayo engxoxweni eqhubekayo, ixhumane nama-API angaphandle, izinqolobane zolwazi nezinhlelo zebhizinisi ukuze iqedele imisebenzi.

Vula inkundla ye-Qianwen AI: Vakashela ku-https://www.qianwenai.com/ bese ungena ku-akhawunti yakho, ungene eMakethe Yamamodeli.
Khetha imodeli oyidingayo: Khetha imodeli efanele ngokwezidingo zakho, i-ASR (ukubhala), i-TTS (ukwenziwa), i-TTS-Next (ukudala), noma i-Realtime (ukusebenzisana ngesikhathi sangempela), bese uchofoza ukuze ungene ekhasini lemininingwane.
Thola i-API Key: Vula isevisi yemodeli kukhonsoli bese udala i-API Key ezosetshenziselwa ukuqinisekisa ukusetshenziswa kwe-interface.
Biza i-interface: Bheka imibhalo ye-API ekhasini lemininingwane yemodeli ngayinye, bese usebenzisa isicelo se-HTTP ukuthumela amapharamitha omsindo noma ombhalo (i-ASR-Next API ayikakhululwa, kufanele ulinde ukuvulwa kwayo).
Hlanganisa nohlelo: Xhuma i-interface kuhlelo lwakho, ku-Agent noma ku-hardware ehlakaniphile, ukuze imisebenzi yezwi iqedwe ku-inthanethi.

Ikhono eliqinile lezilimi eziningi nezilimi zesigodi: Imodeli eyodwa isekela izilimi ezingu-30 nezilimi zesigodi zesiShayina ezingu-16; ukuqaphela izilimi zesigodi ezinzima njenge-Wenzhou kuthuthuke kakhulu.
Ukulungiswa okuhlakaniphile kokubhala: I-ASR isusa ngokwemvelo amagama okugcwalisa nokuphindaphinda, iphinde ihlele kabusha incazelo, ikhiphe umbhalo ohleleke kahle, oshelelayo nonengqondo.
Ukubhala okunembayo ngokwehlukanisa izikhulumi: Ikhipha ndawonye amathegi ezikhulumi, izitempustampu nombhalo isebenzisa indlela yokuphela kuya ekupheleni, futhi idlula ngokuphelele uhlelo oluxhumanisiwe lwe-Fun-ASR.
Ukuqonda umsindo okuthuthukisiwe: I-ASR-Next iyaqonda imizwa, imisindo yemvelo nemisindo yemishini, isuke “embhalweni odluliselwe ekubhalweni” iye “ekuqondeni umsindo ophelele”.
Amakhono okudala ahlanganisiwe: I-TTS-Next ikhiqiza kanye izwi lomuntu, imisindo ekhethekile nemisindo yemvelo, igcine ithoni nemizwa yabalingiswa abaningi ihambisana, ithathe indawo yenqubo yokukhiqiza ngemuva edinga ukwahlukaniswa komsebenzi phakathi kwabantu abaningi.
Ukusebenzisana kwesikhathi sangempela okusezingeni lomuntu: Ilalela futhi ikhulume ngesikhathi esisodwa ngezindlela ezimbili, isekele ukuphazanyiswa, iqaphele imizwa futhi ibize amathuluzi ukuze iqedele imisebenzi.
Ukubambezeleka okuphansi nokusebenza okuphezulu: Impendulo yohlamvu lokuqala ekuqaphelweni okusakazwayo icishe ibe amamilisekhondi angu-160; i-Realtime isebenzisa ukufundiswa othisha abaningi ukuze ihlanganise ukubambezeleka okuphansi namandla okucabanga.

Idemo yokuzama ku-inthanethi: Qwen-Audio-3.1-ASR：https://www.qianwenai.com/models/qwen-audio-3.1-asr-flash
Qwen-Audio-3.1-TTS：https://www.qianwenai.com/models/qwen-audio-3.1-tts-flash
Qwen-Audio-3.1-TTS-Next：https://www.qianwenai.com/models/qwen-audio-3.1-tts-next
Qwen-Audio-3.1-Realtime：https://www.qianwenai.com/models/qwen-audio-3.1-realtime-plus

Amaminithi emihlangano namarekhodi ezingxoxo: Sebenzisa ukubhala kwe-ASR okwehlukanisa izikhulumi nezitempustampu ukuze ukhiqize ngokuzenzakalelayo irekhodi elihlelekile lokuthi “ubani owasho ini futhi nini”. Ukulungiswa kwemvelo kususa ukungcola kwenkulumo, kuthuthukise kakhulu ukusebenza kwemihlangano.
Ukudala ama-podcast nezincwadi ezilalelwayo: Nge-TTS-Next, khiqiza kanye umsindo ophelele wezingxoxo zabantu abaningi, imisindo ekhethekile nemisindo yemvelo. Amathoni nemizwa yabalingiswa abaningi ihlala ihambisana, kususwe isidingo sokuhlukanisa umsebenzi phakathi komethuli, umklami wemisindo nomxubi.
Ukwamukela amakhasimende ngezwi nomsizi ohlakaniphile ngesikhathi sangempela: Ngokusekelwe ekusebenzisaneni ngezindlela ezimbili nasekuqaphelweni kwemizwa kwe-Realtime, yakha isevisi yezwi engaphazanyiswa futhi enozwela, ekwazi nokubiza izinhlelo zebhizinisi phakathi nengxoxo ukuze iqedele imisebenzi efana nokusesha nokufaka ama-oda.
Ukuvakasha kwamazwe ngamazwe nokuxhumana ngezilimi eziningi: Ngokusizakala ngokushintsha ngesikhathi sangempela phakathi kwezilimi ezingu-30 kanye nokwenziwa kwethoni yezwi efanayo kuzo zonke izilimi, inikeza ukuhumusha kwezwi nokuxhumana okushelelayo ekuhleleni uhambo, kwezohwebo lwe-elekthronikhi oluwela imingcele nasemihlanganweni yamazwe ngamazwe.
I-hardware ehlakaniphile ye-AI: Ngokuyihlanganisa ne-QwenNote, izibuko ze-Qianwen AI namanye ama-hardware, abasebenzisi bangethula imisebenzi ngezwi, bangeze izimo futhi balandelele ukuqaliswa kwayo ngaphandle kokuvula ikhompuyutha noma ifoni, benze izwi libe yisango lemvelo le-hardware ye-AI.