Yeqa uye kokuyinhloko
Izifundo ze-AI

Guqula Amavidiyo e-YouTube Abe Ama-Shorts Akhiqizwe I-AI Usebenzisa i-Python

Funda ukufaka nokusebenzisa i-AI Shorts Generator, ithuluzi le-Python elinomthombo ovulekile eliguqula amavidiyo e-YouTube noma asemshinini abe iziqeshana eziqondile, ngokubhala phansi okukhulunywayo, okuvelele okukhethwe i-AI, ukusikwa okuzenzakalelayo, imibhalo engezansi, nama-hook ongawakhetha.

Guqula Amavidiyo e-YouTube Abe Ama-Shorts Nge-Python

Yini i-AI Shorts Generator?

I-AI Shorts Generator iyindlela yamahhala, enekhodi yomthombo evulekile, esikhundleni semikhiqizo ye-SaaS yamavidiyo amafushane efana ne-OpusClip ne-Vidyo.ai. Yamukela i-URL ye-YouTube noma ifayela levidiyo elikukhompyutha, ithole izikhathi ezithembisayo, bese izikhiqiza njengamavidiyo amafushane alungele ukuthunyelwa ngaphandle kwama-watermark noma amakhredithi adingekayo ngaphambi kokusika.

Le phrojekthi ihlanganisa ukubhalwa kwamazwi kwasendaweni nomhlinzeki we-LLM. I-faster-whisper idala umbhalo obhalwe ngezitembu zesikhathi, kanti i-OpenAI, i-Gemini, noma i-MuAPI ihlaziya lowo mbhalo ukuze ibone futhi ihlele iziqeshana ezingaba okuvelele.

Isibonelo sevidiyo emfushane eqondile esicutshunguliwe

Indlela Inqubo Yokucubungula Esebenza Ngayo

  1. Landa noma layisha: Uhlelo lulanda ividiyo ye-YouTube noma lufunde indlela yefayela levidiyo elikukhompyutha.
  2. Bhala amazwi: I-faster-whisper ikhiqiza umbhalo obhalwe ngezitembu zesikhathi endaweni.
  3. Hlukanisa: I-LLM ekhethiwe ibona uhlobo lokuqukethwe nesivinini, njengengxoxo, iphodkasti, isifundo, noma i-vlog.
  4. Hlela okuvelele: I-LLM ihlola izikhathi isebenzisa izimpawu ezifana neziqalo eziqinile, iziqongo zemizwa, izambulo, ukungqubuzana, izitatimende ezifanele ukucashunwa, iziqongo zendaba, nokubaluleka okusebenzayo.
  5. Susa ukuphindaphindeka: Abakhethi abaphambanayo bayasuswa ngokusekelwe kumaphuzu abo.
  6. Khetha iziqeshana: Abakhethi abanamaphuzu aphezulu bayakhethwa ngokuvumelana nenketho ethi --n.
  7. Khiqiza: Okukhethiwe ngakunye kusikwa ngokuzenzakalelayo ukuze kuhambisane nesilinganiso esiceliwe. Ngaphandle kwalapho kukhutshaziwe, i-hook ye-AI eqonda umongo yengezwa ekuqaleni.

Izici Ezibalulekile

  • Ukusekelwa kwe-YouTube namavidiyo asendaweni: Cubungula i-URL yevidiyo noma ifayela elivele ligcinwe kukhompyutha yakho.
  • Ukukhetha okuvelele ngokuzenzakalelayo: Thola izikhathi ezinethuba elikhulu lokuqukethwe kwefomethi emfushane.
  • Ukukhiqiza okuqondile: Khiqiza amavidiyo e-9:16 afanele izinkundla zamavidiyo amafushane.
  • Okuphumayo okuningi: Khiqiza okuvelele okuningi emthonjeni owodwa usebenzisa i---n.
  • Ama-hook akhiqizwa yi-AI: Faka i-hook yokuqala eqonda umongo noma uyikhubaze nge---no-hook.
  • Imibhalo engezansi, ukuhumusha, nokufundwa kwezwi: Le phrojekthi ihlanganisa la makhono okucubungula okuqukethwe ekuhambeni komsebenzi okukodwa.
  • Isixhumi esibonakalayo sewebhu sasendaweni: Faka amavidiyo amaningi emgqeni futhi ulungise izinketho zokukhiqiza ngokubukeka.
  • Ukutholakala kwe-API: I-generator ingafakwa kwamanye amaphrojekthi nge-API yayo.

Izidingo

Ngaphambi kokuqala, qinisekisa ukuthi indawo yakho inalezi zinto:

  • Python 3.10 noma entsha
  • Ukhiye we-API we-OpenAI, Gemini, noma MuAPI
  • Git yokuklona indawo yokugcina
  • Amaphakheji e-Python abhalwe ku-requirements.txt

I-Gemini inikeza isigaba samahhala esinomkhawulo wansuku zonke. I-OpenAI iyakhokhelwa, kanti i-MuAPI inikeza ukufinyelela okukhokhelwayo ngokusetshenziswa ngakunye ngaphandle kokubhalisa. Thola okhiye ku-Google AI Studio, OpenAI Platform, noma MuAPI.

Ukufaka Nokulungiselela

1. Klona indawo yokugcina

git clone https://github.com/pierrenade/short-video-generator-AI.git
cd short-video-generator-AI

2. Dala indawo ebonakalayo

Dala indawo ye-Python ehlukanisiwe ukuze amaphakheji ephrojekthi angaphambanisi nokunye okufakiwe.

python -m venv venv

Ku-Windows, yenza isebenze ngale ndlela:

venv\Scripts\activate

Ku-Linux noma ku-macOS, yenza isebenze ngale ndlela:

source venv/bin/activate

3. Faka okuncikile

pip install -r requirements.txt

4. Lungiselela ifayela lendawo

Dala noma uhlele ifayela le-.env enkombeni yephrojekthi. Setha i-LLM_PROVIDER ibe yi-openai, gemini, noma muapi, bese unikeza ukhiye we-API ohambisanayo.

LLM_PROVIDER=openai

OPENAI_API_KEY=your_openai_key_here
OPENAI_MODEL=gpt-4o-mini

GEMINI_API_KEY=your_gemini_key_here
GEMINI_MODEL=gemini-2.5-flash

MUAPI_API_KEY=your_muapi_key_here

LOCAL_WHISPER_MODEL=base
LOCAL_WHISPER_DEVICE=auto

Udinga kuphela ukhiye we-API osebenzayo womhlinzeki okhethwe yi-LLM_PROVIDER. Izilungiselelo zamamodeli e-OpenAI ne-Gemini aziphoqelekile.

Ukukhetha izilungiselelo ze-Whisper

I-LOCAL_WHISPER_MODEL yamukela i-tiny, base, small, medium, noma large-v3. I-README isebenzisa i-base njengokulungiselela kokuqala.

I-LOCAL_WHISPER_DEVICE yamukela i-auto, cpu, noma cuda. Gcina i-auto ngaphandle uma udinga ukuphoqa ukucubungula nge-CPU noma nge-CUDA.

Khiqiza Ividiyo Yakho Yokuqala Emfushane

Dlulisela i-URL ye-YouTube ku-main.py:

python main.py "https://www.youtube.com/watch?v=video_id"

Lapho ukucubungula sekuqediwe, iziqeshana ezikhiqiziwe zigcinwa enkombeni ye-output.

Umphumela wevidiyo emfushane okhiqizwe yi-AI

Lawula Okuphumayo

Isibonelo esilandelayo sikhiqiza iziqeshana ezintathu eziqondile futhi sicela ukulungiswa kokulandwa kwevidiyo yomthombo okungu-1080:

python main.py "https://www.youtube.com/watch?v=video_id" \
  --n 3 \
  --ratio 9:16 \
  --resolution 1080

Izinketho ezitholakalayo zomugqa womyalo

  • --n: Ibeka ukuthi mangaki ama-clip azokhiqizwa. Okuzenzakalelayo ngu-3.
  • --ratio: Ibeka isilinganiso sobukhulu bomphumela. Okuzenzakalelayo ngu-9:16; sebenzisa 1:1 ukuze uthole umphumela oyisikwele noma unikeze esinye isilinganiso.
  • --resolution: Ibeka ukucaca kokulandwa kwevidiyo yomthombo. Amanani asekelwayo akhonjiswe yiphrojekthi ngu-360, 480, 720, kanye no-1080. Okuzenzakalelayo ngu-720.
  • --language: Iphoqelela ikhodi yolimi ye-Whisper efana no-en. Okuzenzakalelayo kwayo ukuthola ulimi ngokuzenzakalelayo.
  • --no-hook: Ivimbela i-hook ekhiqizwe yi-AI ukuthi ingezwe ekuqaleni kwe-clip ngayinye.

Yenza Ividiyo Yasendaweni

Indlela yefayela lendawo ingasetshenziswa esikhundleni se-URL ye-YouTube. Lesi sibonelo sikhiqiza ama-clip amane aqondile ngokucaca okungu-720 futhi siphoqelela ukubhalwa kombhalo ngesiShayina:

python main.py "/Users/Admin/Folder/video.mp4" \
  --n 4 \
  --ratio 9:16 \
  --resolution 720 \
  --language zh

Sebenzisa indlela ephelele noma ehlobene efanele uhlelo lwakho lokusebenza. Ukufaka indlela phakathi kwezingcaphuno kuyasiza lapho amagama amafolda noma amafayela enezikhala.

Sebenzisa Isixhumi Sobuso Sewebhu Sendawo

Inguqulo yewebhu inikeza indlela ebonakalayo yokubeka amavidiyo amaningi kulayini nokulungisa amafulegi okunikezela. Ukukhethwa komhlinzeki nemininingwane yokufinyelela ye-API kusavela ku-.env; isixhumi siphatha okokufaka kwevidiyo nezinketho zokunikezela.

1. Qalisa i-backend

python3 server.py

2. Nikeza i-frontend

Vula enye i-terminal, uye kufolda yewebhu, bese uqala iseva ye-HTTP emile:

cd web
python3 -m http.server 8000

3. Vula isixhumi

Vakashela http://localhost:8000/shorts-generator-ui.html esipheqululini sakho.

Amasu Athuthukile

Qondanisa isilinganiso sobukhulu nendawo oya kuyo

Sebenzisa --ratio 9:16 kumavidiyo amafushane aqondile ajwayelekile. Sebenzisa --ratio 1:1 lapho udinga i-clip eyisikwele. Iphrojekthi yamukela nezinye izilinganiso.

Phoqelela ulimi lapho ukutholwa kungaqinisekile

Ukutholwa kolimi ngokuzenzakalelayo kuyindlela ezenzakalelayo. Uma usuvele ulwazi ulimi olukhulunywayo, dlulisa ikhodi yalo yolimi ye-Whisper nge---language, efana no---language en.

Khetha inani labazongenela abakhiqizwayo

Okuzenzakalelayo --n 3 kukhiqiza ama-highlight amathathu akhethiwe. Kwandise lapho ufuna abakhethwa abaningi kuvidiyo yomthombo ende, noma kwehlise lapho udinga kuphela umphumela onamandla kakhulu.

Khubaza ama-hook lapho isiqalo sokuqala sanele

I-hook ekhiqiziwe iyinketho. Engeza --no-hook uma ufuna i-clip ngayinye iqale ngokuqondile engxenyeni ekhethiwe yomthombo:

python main.py "https://www.youtube.com/watch?v=video_id" --n 2 --no-hook

Linganisa izinsiza zokubhala umbhalo

Amamodeli e-Whisper atholakalayo aqala ku-tiny aze afike ku-large-v3. Khetha imodeli ehambisana nomshini wakho, futhi usebenzise LOCAL_WHISPER_DEVICE=cuda kuphela lapho kunendawo ye-CUDA ehambisanayo. Kungenjalo, i-auto ivumela uhlelo lokusebenza ukuthi lukhethe idivayisi.

Gcina imininingwane yokufinyelela ingaphandle kwemiyalo

Gcina okhiye bomhlinzeki ku-.env kunokuba ubabeke ngokuqondile emiyalweni ye-terminal noma kukhodi yomthombo. Gwema ukufaka amafayela emvelo agcwele imininingwane lapho wenza i-fork noma unikela kuphrojekthi.

Isiphetho

I-AI Shorts Generator inikeza indlela yasendaweni, enomthombo ovulekile yokuguqula amavidiyo amade e-YouTube noma awasendaweni abe ama-clip amafushane. Ngemva kokulungisa i-Python, umhlinzeki we-LLM, nezilungiselelo zasendaweni ze-Whisper, ungakhiqiza ama-clip amaningana akhethwe ngokuzenzakalelayo ngomugqa womyalo noma usebenzise isixhumi esisekelwe esipheqululini ukuphatha okokufaka nezinketho zokunikezela.

Iphrojekthi iyatholakala ngaphansi kwe-MIT License. Ungabika izinkinga, ufake izinguquko, noma wenze i-fork kuyo kwirepository yayo ye-GitHub.