AI Video Captions

Free, open-source AI video caption generator.
Add trending animated subtitles to any video — 6 styles, word-level animation, 100+ languages.
Self-hosted with Docker. No accounts, no limits.

Try the hosted version

What It Does

Drop in any video and AI Video Captions will:

1. Transcribe it using faster-whisper with word-level timestamps
2. Generate animated ASS subtitles in the style you chose
3. Burn captions directly into the video with FFmpeg
4. Deliver a ready-to-post captioned video in full HD

No accounts. No tracking. No limits. Self-hosted and 100% free.

Screenshots

Upload & Style Picker	Processing Pipeline

Result Viewer	Caption History

Features

• 6 Trending Caption Styles — Hormozi, MrBeast, Karaoke, Minimal, Bounce, Classic — inspired by top-performing short-form content on TikTok, Reels, and Shorts
• Word-Level Animation — Each word animates individually with highlights, wipes, bounces, and scale effects
• 100+ Languages — Automatic language detection with script-aware font fallback (Latin, CJK, Arabic, Devanagari, and more)
• Live Preview — See your chosen caption style on a phone mockup before processing
• Adjustable Position — Drag captions to the exact vertical position you want
• HD Export — CRF 18 quality with original audio preserved
• Job History — Browse all your captioned videos with status and metadata
• 4-Step Pipeline — Visual progress tracking: Upload → Transcribe → Burn → Finalize

Quick Start

Docker (recommended)

git clone https://github.com/nicolaigaina/ai-video-captions.git
cd ai-video-captions
docker compose up

Open http://localhost:3000 and start adding captions.

Local Development

Prerequisites: Python 3.11+, Node.js 20+, FFmpeg

# One-time setup
make setup

# Start both servers
make dev

Frontend runs on localhost:3000, backend on localhost:5000.

Architecture

ai-video-captions/
├── frontend/             Next.js 16 (React 19, Tailwind CSS v4, shadcn/ui)
│   ├── src/app/          Pages: landing, history, result viewer
│   ├── src/components/   UI: dropzone, style picker, phone preview
│   ├── src/actions/      Server actions (proxy to backend API)
│   └── prisma/           SQLite schema for job metadata
│
├── backend/              Flask + faster-whisper + FFmpeg
│   ├── app.py            REST API endpoints
│   ├── caption_job.py    Processing pipeline (transcribe → subtitle → burn)
│   ├── subtitles.py      ASS subtitle generation with animations
│   └── caption_styles.py Style definitions (shared JSON config)
│
├── docker-compose.yml    Production deployment
├── docker-compose.dev.yml  Development with hot reload
└── Makefile              Dev commands

How it works:

1. The Next.js frontend uploads the video to the Flask backend via /api/process
2. The backend queues a job and begins processing in a background thread
3. faster-whisper transcribes the audio with word-level timestamps
4. pysubs2 generates styled ASS subtitles with per-word animations
5. FFmpeg burns the subtitles into the video at the chosen position
6. The frontend polls /api/status/{jobId} and shows real-time progress
7. Once complete, the user previews and downloads the captioned video

Configuration

Copy .env.example to .env and adjust as needed:

Variable	Default	Description
WHISPER_MODEL_SIZE	base	Whisper model: tiny, base, small, medium, large-v3
MAX_FILE_SIZE_MB	500	Maximum upload file size in MB
MAX_DURATION_MINUTES	30	Maximum video duration
MAX_CONCURRENT_JOBS	2	Simultaneous processing jobs
OUTPUT_TTL_HOURS	24	Hours to keep output files before cleanup

Larger whisper models produce better transcriptions but require more RAM and processing time. base is a good default for most use cases.

Tech Stack

Layer	Technology
Frontend	Next.js 16, React 19, TypeScript, Tailwind CSS v4, shadcn/ui
Backend	Python 3.11+, Flask, faster-whisper, pysubs2
Video	FFmpeg (subtitle burn-in, encoding)
Database	SQLite via Prisma
Deployment	Docker Compose

API

Full API documentation is available in docs/API.md.

Method	Endpoint	Description
GET	/api/health	Health check
POST	/api/process	Upload video and start captioning
GET	/api/status/{jobId}	Poll job progress
GET	/api/download/{jobId}	Download captioned video
DELETE	/api/jobs/{jobId}	Delete a job and its files

Development

make dev              # Start frontend + backend
make dev-docker       # Start with Docker (hot reload)
make test             # Run backend tests
make lint             # Lint frontend + backend
make clean            # Remove build artifacts and temp files

See CONTRIBUTING.md for full development setup and guidelines.

License

MIT

---

Built by @nicolaigaina — creator of AutoShorts, AI-powered video repurposing for content creators.

Try AI Video Captions online — no installation required.