README.md

metadata

title: Video Intelligence Chat
emoji: 🎬
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 4.44.1
python_version: '3.10'
app_file: app.py
pinned: false

Video Intelligence Chat

Chat with an AI about any video you upload. The app extracts frames, transcribes speech, and analyzes content using free-tier APIs — all running on Hugging Face Spaces CPU Basic hardware.

---

One-Click Deploy to Hugging Face Spaces

1. Go to huggingface.co/new-space
2. Set SDK to Gradio and Hardware to CPU Basic (free)
3. Clone your new Space and copy all project files in:

git clone https://huggingface.co/spaces/YOUR_USERNAME/YOUR_SPACE_NAME
cd YOUR_SPACE_NAME
cp -r /path/to/video-intelligence-chat/* .
git add .
git commit -m "Initial deploy"
git push

4. Add your secret:
   • Go to Space Settings → Variables and Secrets
   • Click New Secret: NVIDIA_API_KEY = <your key>

The Space rebuilds automatically and is live in ~2 minutes.

---

Getting a Free NVIDIA API Key

1. Sign up at build.nvidia.com (free account)
2. Navigate to any model page and click Get API Key
3. Your free account includes ~1,000 API credits

---

Required Secret

Secret name	Description
NVIDIA_API_KEY	From build.nvidia.com

---

Hardware

• CPU Basic (free tier) is sufficient
• 16 GB RAM, 2 vCPU
• faster-whisper runs locally on Space CPU — no GPU needed

---

Optional Environment Variables

Variable	Default	Description
FRAME_INTERVAL	5	Seconds between extracted frames
WHISPER_MODEL	base	Whisper model size: tiny, base, small
MAX_VIDEO_DURATION	600	Warn if video exceeds this many seconds

---

Recommended Usage

• Video length: Under 5 minutes on the free NVIDIA tier (to conserve API credits)
• Supported formats: MP4, MOV, AVI, WebM
• Max upload size: ~100 MB (Gradio default)

---

Free Tier Limits

Service	Free Allowance
NVIDIA NIM	~1,000 API credits on signup
Hugging Face Spaces	Unlimited CPU Basic spaces
faster-whisper (local)	No limits — runs on Space hardware
HF Space RAM	16 GB RAM, 2 vCPU

---

Tech Stack

Layer	Tool
Frontend / Backend	Gradio (Python)
Hosting	Hugging Face Spaces
Vision AI	NVIDIA NIM — meta/llama-3.2-11b-vision-instruct
Chat LLM	NVIDIA NIM — meta/llama-3.1-8b-instruct
Speech-to-text	faster-whisper (CPU, runs locally)
Video processing	ffmpeg + OpenCV