Qwen3.8-Omni-Flash Uyasebenza

I-Qwen3.8-Omni-Flash iyimodeli yemodi zonke yomdabu eyethulwe yi-Ali Qwen, esekela okokufaka kombhalo, izithombe, umsindo namavidiyo kanye nomongo omude ongu-1M. Igxile emandleni e-Agent “kusukela ekuqondeni kuya ekulethweni”, futhi ingaqedela ngokuzimela izinqubo zomsebenzi eziphelele, ezihlanganisa ukuhlela amavidiyo, ukudala ama-MV, ukuhumusha amadrama amafushane, ukuchaza amafilimu nokwenza izenzo ezilandela amaminithi emihlangano. Uma iqhathaniswa nesizukulwane sangaphambilini, ukusebenza kwayo kukhuphuke ngokwesilinganiso ngaphezu kuka-25%; intengo yokokufaka komsindo nge-API yehle ngaphezu kuka-98%, futhi kwakhishwa izinhlaka ezimbili ezihambisanayo, i-Qwen-MM-Plugins ne-Qwen-Live Harness.

Ukuchaza okulawulwayo komsindo nevidiyo: Umsebenzisi angacacisa ngokukhululeka into echazwayo, ububanzi besikhathi, izinga lemininingwane nefomethi yokuphumayo, ukuze kufezeke “lokho umsebenzisi afuna ukukwazi” kunokuthi “lokho imodeli ekubonile”.
Ukuqonda kwe-Agentic kwamavidiyo amade nomsindo omude: Iqala embuzweni bese inquma ngokuzimela ukuthi ibuke kuphi nokuthi ilalele ini. Ngokusebenzisa imijikelezo eminingi yokuqinisekisa, kusukela ekuhloleni okubanzi kuya kokuningiliziwe, ithola ulwazi olubalulekile, inyuse ukunemba futhi yehlise ukusetshenziswa kwama-token cishe ngo-45.7%.
Ukuqonda nokwenza imihlangano emide: Isekela ngokomdabu okokufaka komsindo nevidiyo okuthatha ihora elilodwa, futhi iqedela kusukela ekuhlukaniseni izikhulumi, ekubhaleni phansi, ekuqondaniseni ubunikazi nasekwenzeni amaminithi kuya ekwenzeni izenzo ezisalindile.
Ucwaningo olujulile lomsindo nevidiyo: Ihlanganisa okuqukethwe kwevidiyo nezidingo zomsebenzisi, icwaninga ngokuzenzakalelayo izinto ezinemodi eziningi, bese ikhiqiza umbiko wocwaningo ogxile kuvidiyo onezithombe nombhalo.
Music2MV: Iqonda ngokuningiliziwe ukwakheka kwengoma, isigqi nemizwa, ikhiphe amagama engoma anezitembu zesikhathi, futhi igcine inqubo ephelele yokudala i-MV, kusukela ekuqondeni umculo kuya ekuhloleni ikhwalithi kwevidiyo yokugcina.
Ukuhumusha amadrama amafushane: Ngomusho owodwa, ibona imigqa yabalingiswa, iyihumushe ngolimi olukhulunywayo, yenze ukudubha ngokukopisha ithoni yezwi, ihlanganise kabusha amathrekhi omsindo futhi ihlole ikhwalithi, ukuze ukulethwa kokuhumusha kuzenzekele.
Ukuchaza amafilimu amade: Ngokufaka ifilimu kanye nesidingo esiwumusho owodwa, iqedela ngokuzenzakalelayo ukuqonda ifilimu yonke, ukufingqa isakhiwo, ukubhala umbhalo wencazelo, ukufaka izwi nomculo, ukuhlela nokunikezela kanye nokuhlola ikhwalithi.
Video2Note: Icindezela amavidiyo athatha amahora amaningi abe amanothi e-PDF ahambisana nezithombe nombhalo futhi anezitembu zesikhathi, bese iwabuyekeza ngokuzenzakalelayo iphinde ilungise amaphutha.
Omni Skill Creator: Ikhipha ama-SOP emibonisweni yokusebenza noma ezifundisweni zochwepheshe, iwaguqule abe ama-Agent Skill angasetshenziswa kabusha futhi aqinisekisiwe.
Ukuprakthiza ukukhuluma ngesikhathi sangempela: Imodela ngokuhlanganyela ukuphimisa nencazelo, iqonde ukuphambuka kwendlela yokukhuluma futhi ikhiqize ngesikhathi sangempela isibonelo sokuphimisa okujwayelekile.

Landela i-WeChat uphendule ngokuthi “开源” ukuze ujoyine iqembu lezingxoxo zamaphrojekthi e-AI avulekile
Ukwakhiwa okuhlanganisiwe kwemodi zonke yomdabu: Imodeli imodela ngokuhlanganyela umbhalo, izithombe, umsindo nevidiyo esakhiweni esisodwa, isebenzisa ukufakwa kwekhodi okuqondanisa amamodi kanye nendawo yokumela ehlanganisiwe ukuze iqonde amamodi ahlukahlukene. Igcina amandla ombhalo acishe afane nawemodeli yombhalo kuphela enosayizi ofanayo; umongo omude wama-token angu-1M uyivumela ukuthi ifake ngokomdabu umsindo nevidiyo okuthatha amahora amaningi, igweme ukuhlukana kolwazi okubangelwa ukucubungula ngezigaba.
Ukuzwa nokuqinisekisa okudingekayo kwe-Agentic: Imodeli iqala ngokuskena okubanzi ukuze ithole izingxenye ezingaba usizo, bese ithola ngokuningiliziwe okuqukethwe okuhlobene komsindo nesithombe ukuze ikuqinisekise. Lokhu kwakha uchungechunge lokuqinisekisa olusuka kokubanzi luye kokuningiliziwe, lugxilise ukubala nesabelomali sama-token ezingxenyeni ezihlobene ngempela.
Ukusebenzisana phakathi komsindo nevidiyo kanye nemvelo ye-Agent: Inselelo enkulu yokwenza ama-Agent omsindo nevidiyo amade ukuthi i-harness ayinakho ukwesekwa komsindo nevidiyo komdabu. Ngakho-ke, i-Qwen ithuthukisa imodeli nochungechunge lwamathuluzi ngokubambisana: i-Qwen-MM-Plugins inikeza ukuzwa okudingekayo, ukubiza amathuluzi namandla okwenza izinqubo zomsebenzi, ixhume nezinhlaka ezidumile zama-Agent ezifana ne-Claude Code ne-OpenClaw, futhi ixhumanise ukuqonda izinto, ukuhlela imisebenzi, ukusebenzisa amathuluzi nokuletha imiphumela kube uchungechunge oluphelele.
Ukuhlonza okuhambisanayo kwezikhulumi eziningi esithombeni nomsindweni: Imodeli imodela ngokuhlanganyela ukusakazwa kwesithombe nomsindo, isebenzise ulwazi olubonakalayo, njengokuvela komuntu esithombeni nokuvuleka komlomo, ukusiza ukuxazulula ukungaqondakali kwezinkomba nezinhlangano emsindweni. Iqedela kusukela ekuhlukaniseni izikhulumi, ekuguquleni inkulumo ibe umbhalo nasekuqondaniseni ubunikazi, okuthuthukisa kakhulu ukunemba ezimweni zemihlangano lapho abantu abaningi bekhuluma ngokushintshana noma amazwi egqagqana.
Ukwakhiwa kokuxhumana okusakazwayo ngesikhathi sangempela: I-Qwen3.8-Omni-Flash-Realtime ithola ukusakazwa komsindo nevidiyo nge-WebSocket/WebRTC, izwe futhi iphendule ngesikhathi sokufakwa. Ukubambezeleka kwe-token yokuqala cishe kungu-600-980ms, kanti i-RTF yokukhiqiza umsindo cishe ingu-0.153; ukulungiswa kokukhuluma kuncike ekumodeleni okuhlangene kokuphimisa nencazelo, imodeli iqhubeka nokuvuselela ukuhlola kwayo lapho kufika inkulumo entsha, ihlukanise amaphutha athinta ukuqonda nokuphimisa kwemvelo.

Ukuyisebenzisa ngokuqondile ekhasini lewebhu: Vula inkundla ye-Qwen AI ku-https://www.qianwenai.com/models/qwen3.8-omni-flash, sesha i-qwen3.8-omni-flash, bese ulayisha ividiyo, umsindo noma isithombe ukuze uxoxe nayo.
Ukubiza nge-API: Faka isicelo se-API Key ye-AliCloud DashScope, usebenzise isixhumi esibonakalayo esihambisana ne-OpenAI, bese uthumela izixhumanisi zezithombe, umsindo namavidiyo kanye nombuzo wombhalo.
Faka ama-plugin ukuze isebenze: Faka i-Qwen-MM-Plugins kumathuluzi afana ne-Claude Code noma i-Qwen Code, bese usho njengokuxoxa ukuthi “@ividiyo.mp4 ngenzele ividiyo yencazelo”; i-AI izoqedela yonke inqubo ngokuzenzakalelayo.

Imodi zonke ehlanganisiwe: Umbhalo, izithombe, umsindo nevidiyo kuphathwa yimodeli eyodwa, esekela umongo omude kakhulu ongu-1M futhi engafaka ngokomdabu ividiyo ethatha ihora.
Kusukela ekuqondeni kuya ekulethweni: Ayigcini nje ngokuqonda umsindo nevidiyo, kodwa ingahlela ngokuzimela, ibize amathuluzi, ikhiqize amavidiyo nemibhalo futhi iqedele imisebenzi ephelele kusukela ekuqaleni kuya ekugcineni.
Amakhono e-Agent avelele: Ihola kakhulu kumabhentshimakhi ama-Agent omsindo nevidiyo; i-WildClawBench-MM inyuke ngamaphuzu angu-36.5 uma iqhathaniswa nesizukulwane sangaphambilini, kanti imisebenzi emide ifinyelele amaphuzu aphezulu angu-69.6 ku-UniClawBench.
Amakhono omsindo adlula i-Gemini: Ukuqashelwa kwezilimi nezinkulumo ezingu-60, izilimi zesigodi zesiShayina ezingu-39 kanye nezinkomba zokuhlonza izikhulumi eziningi konke kudlula i-Gemini 3.8 Flash.
Ukusebenza okuphezulu ngezindleko eziphansi: Ukuqinisekisa kwe-Agentic kwehlisa ukusetshenziswa kwama-token ekuqondeni amavidiyo amade cishe ngo-45.7%, kuyilapho ukunemba kukhuphuka kunokwehla.
Intengo yehle kakhulu: Intengo yokokufaka komsindo nge-API ngehora yehle ngaphezu kuka-98%, kanti yokokufaka komsindo nevidiyo yehle ngaphezu kuka-93%, kunciphisa kakhulu ingcindezi yezindleko ekusetshenzisweni ngobuningi.
Izinto ezivulekile ezihambisanayo ziphelele: I-Qwen-MM-Plugins ne-Qwen-Live Harness zombili zikhishwe njenge-open source futhi ziyahambisana nezinhlaka ezidumile zama-Agent ezifana ne-Claude Code ne-OpenClaw.

Inqolobane ye-GitHub: Qwen-MM-Plugins: https://github.com/QwenLM/Qwen-MM-Plugins
Qwen-Live Harness: https://github.com/QwenLM/Qwen-Live-Harness

Ukukhiqizwa kokuqukethwe kwevidiyo: Faka isidingo esiwumusho owodwa, bese iqedela ngokuzenzakalelayo ukulethwa okuphelele kwencazelo yefilimu, ukudalwa kwe-MV nokuhumusha nokudubha idrama emfushane.
Amathuluzi okuthuthukisa imihlangano: Layisha ukuqoshwa komhlangano othatha ihora, bese ikhiqiza ngokuzenzakalelayo amaminithi nezinto okufanele zenziwe, ize ithumele ama-imeyili futhi idale nemisebenzi.
Ukufunda nokuqongelela ulwazi: Icindezela ngokuzenzakalelayo amavidiyo okufundisa athatha amahora amaningi abe amanothi e-PDF anezithombe-skrini nezitembu zesikhathi.
Ucwaningo olujulile lwevidiyo: Icinga ngokuzenzakalelayo izinto zombhalo nezithombe kuyo yonke iwebhu ezihlobene nokuqukethwe kwevidiyo, bese ikhiqiza umbiko wocwaningo olunzulu onezithombe nombhalo.
Izinsizakalo zokuxhumana ngesikhathi sangempela: Izimo zokuxhumana zomsindo nevidiyo ezinokubambezeleka okuphansi, njengokuphakanyiswa kokukhuluma ngesikhathi sangempela, isevisi yamakhasimende ehlakaniphile nokuzulazula okuthola indawo ngomsindo.