thakurclipforge/clipforge-app

Sabse aasaan tareeka (one-click)

• Windows: START-ClipForge.bat par double-click. Pehli baar Python, FFmpeg aur packages khud install honge, phir browser mein ClipForge khul jayega.
• Mac/Linux: terminal mein ./start-clipforge.sh

ClipForge — apni OpusClip jaisi AI clipping website

Lambi video (YouTube link ya upload) do → AI best moments chunta hai → 9:16 face-tracking crop → animated captions + hook headline → har clip ka virality score (Hook, Flow, Value, Trend) → download.

Features

• YouTube / kisi bhi link se import (yt-dlp) ya file upload (drag & drop)
• Whisper se word-level transcript (Hindi, Hinglish, English, 90+ languages)
• AI moment picker: LLM (OpenAI / Groq / Gemini) ya bina key ke built-in scoring
• "Moments to include" prompt — jaise "jahan paise ki baat ho"
• Clip length: Auto, <30s, 30-60s, 60-90s, 90s-3m · Kitne clips · Time range
• Aspect ratio: 9:16, 1:1, 4:5, 16:9 · Layout: Fill (face tracking) ya Fit (blur background)
• 4 animated caption styles: Karaoke, Bold, Boxed, Minimal (word-by-word highlight)
• Auto hook headline (pehle 3 second title)
• Virality score + sub-scores, AI title, transcript, download, projects history
• Speech cleanup: "umm/uhh" filler words aur lambi khamoshi automatic kat jaati hai
• Brand kit: apna logo (4 positions) + caption/brand colour
• Background music with auto-ducking (aawaz aate hi music dheema)
• Progress bar (neeche chalti line, retention badhati hai) aur Auto zoom (har 4 sec punch-in)
• Har clip ke liye social caption + 5 hashtags, "Copy post" button
• Clip editor: start/end trim, title, caption style/colour badlo → sirf woh clip dobara render
• SRT subtitle download aur "Sab clips ZIP" (clips + SRT + captions_hashtags.txt)

Apne computer par chalao (Windows / Mac / Linux)

1. Install karo: Python 3.10+ aur FFmpeg (Windows: winget install ffmpeg)
2. Folder mein terminal kholo:

   pip install -r requirements.txt
   copy .env.example .env      (Mac/Linux: cp .env.example .env)
   uvicorn app.main:app --port 8000
   

3. Browser mein kholo: http://localhost:8000

Pehli baar Whisper model (~250 MB "small") khud download hoga.

Behtar AI ke liye (recommended)

.env mein LLM_API_KEY daalo. Groq ka free key (console.groq.com) bhi kaafi hai.

Hindi captions

Devanagari ke liye Noto Sans Devanagari font ki .ttf file fonts/ folder mein daalo aur app/captions.py mein font ka naam "Noto Sans Devanagari" kar do.

Internet par live karna

• Docker: docker build -t clipforge . && docker run -p 8000:8000 --env-file .env clipforge
• Sasta server: Hetzner / DigitalOcean / AWS Lightsail (4 GB RAM+), ya GPU ke liye RunPod
• Domain jodna: Cloudflare + Nginx reverse proxy
• Note: YouTube kuch cloud servers ko block karta hai; tab yt-dlp ke saath cookies use karo ya users se file upload karwao.

Aage kya jodna hai (roadmap)

Login + credits (Supabase/Clerk + Razorpay), clip editor (trim, caption edit), filler word removal, B-roll (Pexels API), dubbing (ElevenLabs), YouTube/Instagram auto-post, team workspace.

Code ka naksha

• app/main.py — web server, job queue, API
• app/pipeline.py — download → transcript → select → render
• app/selector.py — AI moment picker + virality score
• app/reframe.py — face tracking smart crop
• app/captions.py — animated captions (ASS)
• static/index.html — poori website ka UI