Guqula Amavidiyo e-YouTube Abe Ama-Shorts Akhiqizwe yi-AI Nge-Python

Yini i-AI Shorts Generator?

I-AI Shorts Generator iphrojekthi yamahhala, enekhodi evulekile ye-Python yokuguqula amavidiyo amade e-YouTube noma amafayela evidiyo asendaweni abe iziqeshana ezimfushane ezilungele ukuthunyelwa. Ibhala ngokuzenzakalelayo umbhalo walokho okukhulunywayo emthonjeni, ikhombe izikhathi ezingase zihehe, isuse okukhethiwe okudlulanayo, bese ikhiqiza abakhethiwe abangcono kakhulu ngesilinganiso esime mpo noma esenziwe ngokwezifiso.

Iphrojekthi ibekwe njengenye indlela ezisingathwa ngumnikazi wayo kunemikhiqizo ye-SaaS yamavidiyo amafushane. Ayifaki amakhredithi noma ama-watermark ngaphambi kwesiqeshana. Izici zayo ezibanzi zihlanganisa nemibhalo engezansi, ukuhumusha, izwi eliqoshiwe, ama-hook okuvula aqonda umongo, isixhumi esibonakalayo somugqa womyalo, isixhumi esibonakalayo sewebhu sasendaweni, kanye ne-API yokuyihlanganisa namanye amaphrojekthi.

Isibonelo somphumela

Iziqeshana ezikhiqiziwe zingaguqula izinkulumo, izingxoxo, ama-podcast, izifundo, kanye nezinto ezifanayo zefomethi ende zibe amavidiyo alungele omakhalekhukhwini.

Indlela Esebenza Ngayo Inqubo Yokucubungula

1. Ukuthola umthombo: Uhlelo ludawuniloda ividiyo ye-YouTube noma lufunde indlela yefayela lasendaweni oyinikezayo.
2. Ukubhala okukhulunywayo: faster-whisper idala umbhalo obhalwe ngezitembu zesikhathi endaweni. Lesi sinyathelo siyefana kubo bonke abahlinzeki be-LLM abasekelwayo.
3. Ukuhlukanisa okuqukethwe: I-LLM ekhethiwe inquma ukuthi umthombo ufana yini ne-podcast, ingxoxo, isifundo, i-vlog, noma olunye uhlobo lokuqukethwe. Iphinde ihlole ijubane lokugeleza ukuze ukukhethwa kwezithombe ezigqamile kushintshwe ngokuya ngento esetshenziswayo.
4. Ukubeka izithombe ezigqamile ezingeni: I-OpenAI, i-Gemini, noma i-MuAPI ihlola umbhalo ukuze ithole ama-hook, iziqongo zemizwa, imibono eqinile, izambulo, ukungqubuzana, izitatimende ezingacashunwa, iziqongo zendaba, kanye nenani elisebenzisekayo. Ibuyisa abakhethiwe bezithombe ezigqamile abahlelwe ngokwamazinga, abanamaphuzu asuka ku-0 kuya ku-100.
5. Ukususa ukuphindaphindeka: Abakhethiwe abadlulanayo bahlanganiswa ngokwamaphuzu abo.
6. Ukukhetha abaphezulu ku-N: Abakhethiwe abanamaphuzu aphezulu bayakhethwa kusetshenziswa inketho ethi --n.
7. Ukukhiqiza: Isigaba ngasinye esikhethiwe sinqunywa ngokuzenzakalelayo bese sikhiqizwa ngesilinganiso esiceliwe. Ngaphandle kokuthi kukhutshazwe, i-hook ekhiqizwe yi-AI eqondene nomongo ibekwa ekuqaleni.

Izidingo

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

I-Gemini inikeza izinga lamahhala elinomkhawulo wansuku zonke. I-OpenAI iyakhokhelwa, kanti i-MuAPI isebenzisa ukufinyelela okukhokhelwayo ngokusetshenziswa ngakunye ngaphandle kokubhalisa. Imigomo nezintengo zabahlinzeki zingashintsha, ngakho hlola isevisi efanele ngaphambi kokucubungula iqoqo elikhulu.

Faka Iphrojekthi

1. Klona i-repository

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

2. Dala indawo ebonakalayo

Ku-Windows, sebenzisa:

python -m venv venv
venv\Scripts\activate

Ku-Linux noma ku-macOS, sebenzisa:

python -m venv venv
source venv/bin/activate

Indawo ebonakalayo igcina amaphakheji e-generator ehlukene kwamanye amaphrojekthi e-Python akukhompyutha yakho.

3. Faka okuncike kukho

pip install -r requirements.txt

Lungiselela i-LLM ne-Whisper

Dala ifayela le-.env kumkhombandlela wephrojekthi. Setha i-LLM_PROVIDER ibe yi-openai, gemini, noma muapi, bese wengeza 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 ovumelekile womhlinzeki okhethwe yi-LLM_PROVIDER. Izinkambu zamamodeli eziboniswe ze-OpenAI ne-Gemini ziyazikhethela.

• Gemini: Thola ukhiye ku-Google AI Studio.
• OpenAI: Thola ukhiye ku-OpenAI Platform.
• MuAPI: Thola ukhiye ku-MuAPI.

Khetha imodeli ye-Whisper

I-LOCAL_WHISPER_MODEL yamukela i-tiny, base, small, medium, noma large-v3. Isibonelo esizenzakalelayo sisebenzisa i-base. Ekukhetheni idivayisi, sebenzisa i-auto, cpu, noma cuda.

Shiya i-LOCAL_WHISPER_DEVICE=auto uma ufuna uhlelo lukhethe idivayisi etholakalayo. Sebenzisa i-cuda uma unendawo ehambisanayo ye-CUDA, noma i-cpu uma ufuna ngokusobala ukucubungula nge-CPU.

Khiqiza Isiqeshana Sakho Sokuqala

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

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

Ngokuzenzakalelayo, uhlelo lukhiqiza iziqeshana ezintathu ngesilinganiso esingu-9:16 futhi lusebenzisa ukulungiswa kokudawuniloda komthombo okungu-720. Amafayela akhiqiziwe agcinwa kumkhombandlela we-output.

Lawula Inani, Isilinganiso, Nokulungiswa

Umyalo olandelayo ukhetha izithombe ezigqamile ezintathu, uzifomethe njengamavidiyo ame mpo, bese ucela ukulungiswa komthombo okungu-1080:

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

• --n ilawula ukuthi zingaki iziqeshana ezikhiqizwayo. Okuzenzakalelayo ngu-3.
• --ratio yamukela isilinganiso esenziwe ngokwezifiso. Sebenzisa 9:16 kumavidiyo amafushane ame mpo noma 1:1 kokuphumayo okuyisikwele.
• --resolution ilawula ukulungiswa kokudawuniloda kwevidiyo yomthombo. Amanani abhaliwe asekelwayo ngu-360, 480, 720, kanye no-1080.

Qalisa Ividiyo Yasendaweni

Indlela yefayela lendawo ingathatha indawo ye-URL ye-YouTube. Lesi sibonelo sakha iziqeshana ezine ezime mpo ngokulungiswa okungu-720 futhi siphoqa i-Whisper ukuthi ithathe inkulumo njengesiShayina:

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

Ukusethwa kolimi okuzenzakalelayo ngu-auto. Sebenzisa --language ngekhodi yolimi ye-Whisper efana no-en lapho ukutholwa okuzenzakalelayo kungafaneleki noma usuvele ulwazi ulimi olukhulunywa kuvidiyo.

Khubaza I-Hook Ekhiqiziwe

Ama-hook akhiqizwa yi-AI avulwa ngokuzenzakalelayo. Ukuze ukhiqize okuvelele ngaphandle kokufaka i-hook entsha ekuqaleni, engeza --no-hook:

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

Lokhu kuyasiza lapho ingxenye ekhethiwe isivele inesiqalo esinamandla noma lapho ufuna ividiyo emfushane iqale ngqo ngendlela evela ngayo emthonjeni.

Sebenzisa Isixhumi Esibonakalayo Sewebhu Sasendaweni

Inguqulo yewebhu inikeza indlela ebonakalayo yokubeka amavidiyo amaningi emgqeni nokulungisa amafulegi okunikezela. Isafunda umhlinzeki we-LLM nemininingwane ye-API ku-.env; isixhumi esibonakalayo siphatha kuphela okokufaka kwevidiyo nezinketho zokunikezela.

1. Qalisa i-backend

python3 server.py

2. Nikeza i-frontend

Vula enye i-terminal kumkhombandlela wephrojekthi bese usebenzisa:

cd web
python3 -m http.server 8000

3. Vula isixhumi esibonakalayo

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

Amathiphu Okusetshenziswa Okuthuthukile

Qondanisa isilinganiso nobabili nendawo oya kuyo

Sebenzisa 9:16 kumavidiyo amafushane ajwayelekile ame mpo. Khetha 1:1 lapho udinga okuphumayo okuyisikwele, noma unikeze esinye isilinganiso esisekelwa umkhiqizi ukuze uthole ukwakheka okuhlukile.

Nciphisa iziqeshana ngesikhathi sokuhlola

Setha --n 1 ngesikhathi sokucushwa kokuqala. Lokhu kunciphisa inani leziqeshana ezikhiqizwayo ngenkathi uqinisekisa imininingwane ye-API, indlela yokubhala inkulumo, izilungiselelo zolimi, ama-hook, nobukhulu bokuphumayo.

python main.py "https://www.youtube.com/watch?v=video_id" --n 1 --resolution 360

Phoqa ulimi lapho kufanele

Ukutholwa kolimi okuzenzakalelayo kulula, kodwa ukusetha ngokusobala --language kungasiza lapho ulwazi ulimi lomthombo. Le nketho isebenza ekubhalweni kwenkulumo nge-Whisper.

Lungisa izilungiselelo zokubhala inkulumo zasendaweni

Ukubhala inkulumo kwenziwa endaweni nakuba ukuhlaziywa kokuvelele kusebenzisa umhlinzeki wangaphandle we-LLM. Khetha imodeli ye-Whisper nedivayisi ku-.env ngokwezinsiza ezitholakala emshinini wakho. I-README iveza lezi zilungiselelo kodwa ayinqumi imodeli eyodwa yazo zonke izinhlobo zomsebenzi.

Gcina ukucushwa komhlinzeki kuhlukile ezinqumweni zesixhumi esibonakalayo

Ukushintsha ku-CLI uye kusixhumi esibonakalayo sewebhu akushintshi umhlinzeki we-LLM omisiwe. Buyekeza LLM_PROVIDER kanye nokhiye we-API ohambisanayo ku-.env noma nini lapho ushintsha phakathi kwe-OpenAI, i-Gemini, ne-MuAPI.

Buyekeza iziqeshana ezikhiqiziwe ngaphambi kokushicilela

Uhlelo lokugqamisa lusebenzisa ukuhlaziywa kombhalo obhaliwe nohlaka lokusabalala ukuze lukhethe ngokuzenzakalelayo abazokhethwa. Buyekeza amafayela aku-output ukuze uqinisekise ukuthi ukunqampuna, ukubhala inkulumo, umongo, ne-hook ongayikhetha kufanelana nokusetshenziswa okuhlosiwe ngaphambi kokuwathumela.

Isiphetho

AI Shorts Generator ihlanganisa ukubhala inkulumo kwasendaweni nge-Whisper, ukutholwa kokuvelele okusekelwe ku-LLM, ukususa okuphindaphindiwe, ukukala, nokunqampuna okuzenzakalelayo kube ukugeleza komsebenzi okukodwa komthombo ovulekile. Ngemva kokufaka okuncikile nokucushwa komhlinzeki ku-.env, ungacubungula i-URL ye-YouTube noma ifayela lendawo ngomugqa womyalo. Isixhumi esibonakalayo sewebhu sasendaweni sinikeza enye indlela lapho ufuna ukubeka amavidiyo emgqeni nokukhetha izinketho zokunikezela ngokubukeka.

Iphrojekthi itholakala ngaphansi kweLayisensi ye-MIT. Ungabika izinkinga, ufake isandla ngezinguquko, noma wenze i-fork yendawo yokugcina ku-GitHub.