Isikhathi sedatha: Uhlu lwanyanga zonke lwe-GitHub Trending lukaJanuwari 2026｜Ukwanda kwezinkanyezi +18.4k｜Isilinganiso sansuku zonke 600+ ⭐
1. Ukwanda Okukhulu Kwezinkanyezi Nezinselele Zomkhakha
Ohlwini lwanyanga zonke lwe-GitHub lukaJanuwari 2026, i-ViralCut ingene kwabathathu abaphezulu ngokwengeza izinkanyezi ezingu-18,000 ngenyanga eyodwa. Uma ubheka ijika lokukhula kwayo ezinsukwini ezingu-90 ezedlule, licishe libe umugqa okhuphuka uqonde phezulu—yakhula ngokushesha kusukela yethulwa, ngaphandle kokusebenzisa imali ekukhangiseni, ngenxa yokwabelana ngomlomo phakathi komphakathi wonjiniyela.
Lokhu kubambe izinkinga ezintathu ezazilokhu ziqoqeka embonini yabadali bokuqukethwe:
Izindleko ezifihlekile zokusabalalisa ezinkundleni eziningi. Ividiyo efanayo kufanele ilungiswe ngokuhlukile ibe eye-Douyin, Bilibili, WeChat Channels ne-YouTube Shorts. Inkundla ngayinye inezidingo ezihlukile zosayizi, bitrate nesembozo; ukuhamba komsebenzi okujwayelekile kwe-PR kungakhiqiza amavidiyo angu-3-5 kuphela ngobusuku.
Ukuhlukana kwamathuluzi okukhiqiza amavidiyo nge-AI. I-Sora, i-Kling, i-Jimeng ne-Pika ngayinye ine-API yayo; ukubhala, ukuqopha izwi, ukuhlela nemibhalo engezansi kusakazeke kumathuluzi ahlukene, idatha ayixhumani, futhi ukusetshenziswa kwama-token namahora okusebenza akukwazi ukubalwa kahle.
I-ecosystem ye-MCP “inephrothokholi kodwa ayinakho ukuhlelwa”. Nakuba i-Model Context Protocol ixazulule inkinga yokuxhawulana phakathi kwamamodeli amakhulu nezinsiza, bekungekho mhleli wokuhamba komsebenzi owakwazi ukuhlanganisa “ukukhetha isihloko → ukukhiqiza → ukuhlela → ukusabalalisa → ukubuyisa idatha” kube uchungechunge olulodwa oluphelele.
I-ViralCut ibekwe ngqo kulezi zindawo ezintathu ezihlangana kuzo. Izichaza njenge-pipeline yokuzenzakalela yamavidiyo amafushane esekelwe ku-MCP—ihlanganisa ukukhiqiza nge-AI, ubunjiniyela bokuhlela nokusebenza kokusabalalisa kube ukuhamba komsebenzi okuhlelekayo.
2. Isingeniso Sephrojekthi: Iyini i-ViralCut?
I-ViralCut iwuhlaka oluvulekile lokuzenzakalela kwamavidiyo amafushane olwakhiwe nge-MCP (Model Context Protocol). Umhleli oyinhloko ubhalwe nge-Go, ingxenye yangaphambili isekelwe ku-Next.js, kanti ingxenye yokuhlela ihlanganisa i-FFmpeg ne-Remotion. Iphrojekthi yavulwa ngoNovemba 2025 yithimba langaphambili labakhi bezakhiwo zamavidiyo bakwa-ByteDance, futhi ilayisense yayo yi-Apache-2.0.
Isimangalo sayo esiyinhloko siqondile: ukunikeza i-AI Agent ukukhiqizwa kwamavidiyo amafushane adumile ngendlela yezimboni, kuyilapho izinqumo zobuciko zishiyelwa abantu.
Ayiyona imodeli ye-AI ekhiqiza yonke into kusukela ekuqaleni kuze kube sekugcineni, futhi ayisona isofthiwe yokuhlela kuphela; iyisendlalelo sokuhlela nokuxhumanisa. Amakhono ayo wonke axhunywa ngendlela yama-MCP Server, kanti i-ViralCut inikeza ngaphezulu ukuhlelwa kwemisebenzi okuhlanganisiwe, umshini wezimo, ukuqapha nendlela yokubuyela emuva.
Ekupheleni kukaJanuwari 2026, indawo yokugcina ikhodi yayisivele inama-adapter angu-47 e-MCP Server avela ngokusemthethweni nasemphakathini, ahlanganisa amamodeli amakhulu amavidiyo e-AI, ama-LLM okubhala, i-TTS yezwi, izinjini zokuhlela nama-API okusabalalisa.
3. Imisebenzi Eyinhloko
1. Injini yokuhlela yemvelo ye-MCP
Incazelo yokuhamba komsebenzi esekelwe ku-DAG; i-YAML ne-JSON kungahlelwa nhlangothi zombili.
Inezindawo ezingu-12 ezisetshenziswa njalo: ukubiza i-LLM, ukukhiqiza ividiyo, ukuhlanganisa izwi, ukukhiqiza imibhalo engezansi, ukukhiqiza isembozo, ukuhlanganisa ukuhlela, ukusabalalisa ezinkundleni, ukubuyisa idatha, ukuhlola i-A/B, ukuhlaziya izici zevidiyo edumile, ukuhlola amagama avinjelwe kanye ne-watermark yamalungelo obunikazi.
Isekela ukuphinda kuzanywe lapho kwehluleka, ukuqhubeka kusukela lapho kumiswe khona, ukunciphisa imisebenzi ehambisanayo kanye nesabelomali sezindleko.
2. Ukunikezwa okuzenzakalelayo ezinkundleni eziningi
Ikhipha ngokuchofoza okukodwa amafomethi angu-9:16, 1:1, 16:9 no-3:4.
Ihlanganisa amapharamitha anconyiwe wokufaka ikhodi ezinkundleni ezahlukene (H.264/H.265, bitrate nesikhawu sama-keyframe).
Ikhiqiza ngokuzenzakalelayo izembozo ezihambisana nezindinganiso zenkundla, izama ukunciphisa ubungozi bokwenqatshwa ekuhlolweni.
3. Imodi yokubambisana yama-AI Agent
Ihlinzeka ngezinhlobo ezintathu zama-Agent akhelwe ngaphakathi:
planner,editorkanyereviewer.
I-Planner ihlukanisa isihloko sibe yimisebenzi emincane engenziwa, i-Editor ihlela amathuluzi ukuze iqede ukukhiqiza, kanti i-Reviewer inikeza amaphuzu ezici zevidiyo edumile (isibikezelo sezinga lokuqeda ukubuka, ijika lemizwa namandla e-Hook).
Umsebenzisi angabhala ngolimi lwemvelo ukuze ashintshe izinqumo ze-Agent nganoma yisiphi isikhathi.
4. Ukubuyiswa kwedatha nokufunda kumavidiyo adumile
Ixhuma ne-Feishu, i-Notion ne-Slack, bese ibuyisa ngokuzenzakalelayo idatha yokusabalalisa.
Ihlanganisa imojuli ye-RAG eguqula amavidiyo adumile omlando abe yisizindalwazi sama-vector, bese isebenzisa lokho ukuthuthukisa i-Prompt yokukhiqiza elandelayo.
Isekela ukufakwa ngaphakathi kwendawo yangasese, ukuze idatha ingaphumi esizindeni.
5. Ukubonakala nokulawula izindleko
I-Web UI ikhombisa ngesikhathi sangempela isikhathi esisetshenziswe yinodi ngayinye, ukusetshenziswa kwama-Token kanye nezindleko zokubiza ama-API.
Isekela ukunqanyulwa okuzenzakalelayo lapho kufinyelelwa kwisabelomali kanye nokwehliswa kwezinga lemodeli (isibonelo, ngemva kokuthi i-GPT-4o isisebenzise sonke isabelomali, ishintshela ngokuzenzakalelayo ku-4o-mini).
4. Imiyalo Yokufaka Ngokuchofoza Okukodwa
I-ViralCut inikeza izindlela ezimbili zokufaka: i-Docker Compose ne-Kubernetes, futhi ingaqalwa ngokushesha ngemizuzu engu-3:
Indlela yokuqala: Docker Compose (inconyelwe onjiniyela abazimele)
# Klona indawo yokugcina ikhodi git clone https://github.com/viralcut/viralcut.git cd viralcut # Kopisha isifanekiso sokucushwa cp .env.example .env # Hlela i-.env, ufake OPENAI_API_KEY, DOUYIN_TOKEN, BILIBILI_TOKEN njll. # Qalisa ngokuchofoza okukodwa docker compose up -d # Buka i-Web UI open http://localhost:3000 Indlela yesibili: Kubernetes (inconyelwe amaqembu)
helm repo add viralcut https://charts.viralcut.dev helm install my-viralcut viralcut/viralcut \ --namespace viralcut \ --create-namespace \ --values values.yaml ⚠️ Ekuqaliseni kokuqala kufanele usebenzise
viralctl initukuqedela ukubhaliswa kwe-MCP Server nokugunyazwa kwe-OAuth yezinkundla.
5. Uchungechunge Lokuxhumanisa Amathuluzi
Amandla angempela e-ViralCut ukuthi ihlanganisa uchungechunge lwamathuluzi e-AI abekade ehlukene lube “umugqa wokukhiqiza wezimboni” owodwa:
[Ugqozi lwesihloko] → ChatGPT/Claude (MCP) ↓ [Ukukhiqiza umbhalo] → LLM Node ↓ [Iskripthi se-storyboard] → LLM Node + Isizindalwazi sezici zamavidiyo adumile (RAG) ↓ [Ukukhiqiza izithombe] → Sora / Kling / Jimeng (MCP Adapter) ↓ [Ukuhlanganisa izwi] → Volcano Voice / ElevenLabs (MCP Adapter) ↓ [Ukukhiqiza imibhalo engezansi] → Whisper / Doubao ASR (MCP Adapter) ↓ [Ukuhlela nokuhlanganisa] → FFmpeg / Remotion Node ↓ [Ukusabalalisa ezinkundleni] → Douyin / Bilibili / WeChat Channels / YouTube (MCP Adapter) ↓ [Ukubuyisa idatha] → Feishu Bitable / Notion Database ↓ [Ukufunda kumavidiyo adumile] → Isizindalwazi sama-vector sithuthukisa i-Prompt elandelayo Lonke lolu chungechungelusebenzisa ukuxhumana kwephrothokholi ye-MCP, futhi noma iyiphi inodi ingashintshwa, ithuthukiswe noma ifakwe endaweni yangasese. Isibonelo, uma ufuna ukushintsha ukukhiqizwa kwezithombe kube ukuhamba komsebenzi kwe-ComfyUI ozakhele wena, usebenzise i-SoVITS yasendaweni ekukhiqizeni izwi, noma ugcine kuphela i-App yenkampani ekusabalaliseni ezinkundleni, konke kuyenzeka.
Lokhu “kuxhuma nokukhipha ezingeni lephrothokholi” kuyinto enkulu ehlukanisa i-ViralCut kumathuluzi e-SaaS ajwayelekile—youke awusaboshiwe kumhlinzeki oyedwa; uma ithuluzi lingasebenzi kahle, ungalishintsha.
6. Ukuqhathanisa Nezimbangi
Isici｜ViralCut｜Ukuzenzakalela kwe-Jianying AI｜n8n + Ama-plugin e-AI｜Coze
Izinga lephrothokholi｜MCP yemvelo｜I-API yangasese｜HTTP Webhook｜Iphrothokholi yangasese
Izinga lokuvuleka｜Ivuleke ngokuphelele (Apache-2.0)｜Ivaliwe｜Umthombo ovulekile (Fair-code)｜Ivaliwe
Indlela yokufaka｜Iseva yakho / Ifu langasese｜I-SaaS kuphela｜Iseva yakho / SaaS｜I-SaaS kuphela
Ukuxhumana namamodeli evidiyo e-AI｜12+ asemthethweni, 35+ omphakathi｜I-ByteDance kuphela｜Kuncike kuma-plugin omphakathi｜I-ByteDance kuphela
Ukusabalalisa ezinkundleni eziningi｜Douyin/Bilibili/WeChat Channels/YouTube/X｜I-Douyin kuphela｜Noma yikuphi (kudinga ukuzakhela)｜I-Douyin kuphela
Ubumfihlo bedatha｜✅ Kusekelwa ngokuphelele｜❌｜✅｜❌
Umjikelezo wokufunda kumavidiyo adumile｜✅ I-RAG eyakhelwe ngaphakathi｜❌｜❌｜Kusekelwa ngokwengxenye
Ubunzima bokufunda｜Phakathi (kudingeka wazi i-MCP)｜Phansi｜Phezulu｜Phansi
Njengoba kubonakala, umgoqo oyinhloko we-ViralCut ukuhlanganisa “ukuhlela ukuhamba komsebenzi we-AI” kanye “nokukhiqizwa kwamavidiyo amafushane ngendlela yezimboni” kube isixazululo esiqondile, bese ibuyisela ngokuphelele amandla ezinqumo kumsebenzisi. Emakethe ephethwe yi-SaaS evaliwe, lesi sikhundla siyivelakancane kakhulu.
7. Abasebenzisi Abafanele
Onjiniyela be-full-stack / backend: Bafuna ukusebenzisa i-MCP ukuhlanganisa amakhono e-AI akhona abe umkhiqizo, kodwa badinga uhlaka lokuhlela olubonakalayo noluhlelekayo.
Amaqembu abadali bokuqukethwe / izikhungo ze-MCN: Kufanele bakhiqize amavidiyo angaphezu kuka-50 ngosuku, futhi bakhathele ukuhlela ngesandla nomsebenzi ophindaphindwayo.
Osomabhizinisi bezinhlelo ze-AI: Bafuna ukuqinisekisa ngokushesha i-MVP yomkhakha “AI + amavidiyo amafushane”, ngaphandle kokuboshelwa ku-API yenkampani eyodwa.
Abaphathi bezinhlelo / SRE: Amabhizinisi anezidingo zokucubungula idatha yemultimedia futhi adinga iphayiphi elingafakwa endaweni yangasese.
Onjiniyela abazimele: Umuntu oyedwa ufuna ukukhiqiza ividiyo edumile, agxilise amandla akhe ebucikweni kunemininingwane yobunjiniyela.
8. Ukuthobela Nemvume Yezimpawu Zokuhweba
Amagama e-Sora, Kling, Jimeng, Pika, ElevenLabs, Whisper, Doubao, SoVITS, Jianying, Coze, Douyin, Bilibili, WeChat Channels, YouTube, Slack, Notion, Feishu, FFmpeg, Remotion, ChatGPT, Claude, ComfyUI, n8n, GPT-4o ashiwo kulo mbhalo ayizimpawu zokuhweba ezibhalisiwe noma amagama emikhiqizo abanikazi bawo, futhi asetshenziswe lapha kuphela ukuchaza amakhono obuchwepheshe; lokhu akwakhi ukumela, ukugunyaza noma ubudlelwano bokubambisana kwezentengiselwano nganoma iyiphi indlela. I-ViralCut, njengephrojekthi evulekile, ilandela ilayisense ye-Apache-2.0 futhi ayinabudlelwano bokuba ngaphansi kwabanikazi balezi zimpawu. Lapho usebenzisa ama-API nezinsiza ezihlobene endaweni yokukhiqiza, sicela uzibophezele ngokwakho ekuthobeleni imigomo yesevisi, izidingo zokuthobela idatha kanye nemithetho yamalungelo obunikazi yezinkundla ezifanele.
9. Iseluleko Sokuqala Nesifinyezo
Uma uhlela ukuhlanganisa uchungechunge lwamathuluzi emkhakheni we-“AI + okuqukethwe” ngo-2026, i-ViralCut cishe iyonakuphela iphrojekthi evulekile ku-GitHub okwamanje ehlanganisa kahle i-MCP, ubunjiniyela bamavidiyo nokusebenza kokusabalalisa. Ukuduma kwayo akulona ithuba—empeleni kubambe ukuhlangana kwamathuba amathathu esikhathi: “ukuhlukana kwamathuluzi e-AI × ukukhiqizwa kwabezindaba ngendlela yezimboni × ukuvuthwa kwe-ecosystem ye-MCP”.
Indlela enconyiwe yokuqala: qala ngokusebenzisa i-Docker Compose ukuze uqhube ukuhamba komsebenzi kwesibonelo → xhuma amamodeli e-AI akhona → guqula i-SOP yakho ibe yi-DAG → xhuma ama-akhawunti ezinkundla zakho → zakha isizindalwazi sama-vector samavidiyo adumile. Ezinsukwini ezintathu, ungakha “ifekthri yakho yokuzenzakalela kwamavidiyo amafushane”.
Ikheli lendawo yokugcina ikhodi ku-GitHub: https://github.com/viralcut/viralcut