Subvision is an AI-powered subtitle generator. Upload your video, and Subvision will automatically generate subtitles using AI, track your process, and let you download the result when it's ready.

• Video upload with resumable uploads (tus protocol)
• Automatic subtitle generation using AI
• Track processing status
• Download processed videos with subtitles
• Modern web UI (Next.js, Tailwind CSS)
• Backend with Gin, MinIO, RabbitMQ, tusd

• Client: Next.js app (client/)
• Server: Go backend (server/)
• Vfx: Node.js service for subtitle rendering and effects using Remotion (vfx/)
• MinIO: S3-compatible object storage for uploads
• RabbitMQ: Job queue for processing
• tusd: Resumable upload server

flowchart TD
    A[Client - Next.js] -- Upload video via tusd --> B[Server - Go, tusd handler]
    B -- Store video --> C[MinIO]
    C -- Upload complete event --> D[Server - Publish upload_jobs queue]
    D -- upload_jobs --> E[Server - ProcessUploadedFile]
    E -- Download video from MinIO<br>Convert to WAV<br>Transcribe with whisper.cpp --> F[Transcription Segments]
    E -- Publish generate_vfx_jobs --> G[VFX Service - Node.js, Remotion]
    G -- Download video from MinIO<br>Generate frames<br>Combine with ffmpeg --> H[Processed Video]
    G -- Upload processed video --> I[MinIO outputs/id]
    I -- Ready for download --> A

    style A fill:#e0f7fa,stroke:#0097a7
    style B fill:#fffde7,stroke:#fbc02d
    style C fill:#e8f5e9,stroke:#388e3c
    style D fill:#f3e5f5,stroke:#8e24aa
    style E fill:#fff3e0,stroke:#f57c00
    style F fill:#fce4ec,stroke:#d81b60
    style G fill:#e1f5fe,stroke:#0288d1
    style H fill:#f9fbe7,stroke:#afb42b
    style I fill:#e8f5e9,stroke:#388e3c

• Speech-to-text is powered by whisper.cpp, a fast and portable implementation of OpenAI's Whisper model.

• Docker & Docker Compose

git clone --recurse-submodules https://github.com/yourusername/subvision.git
cd subvision

Copy the example environment files and edit them as needed:

cp server/.env.example server/.env
cp client/.env.example client/.env
cp vfx/.env.example vfx/.env

• Edit server/.env for backend configuration (see server/.env.example)
• Edit client/.env for frontend configuration (see client/.env.example)
• Edit vfx/.env for VFX service configuration (see vfx/.env.example)

docker-compose up --build

• Frontend: http://localhost:3000
• Backend API: http://localhost:8080
• MinIO Console: http://localhost:9001 (user: minio, pass: minio123)
• RabbitMQ Console: http://localhost:15672 (user: guest, pass: guest)

cd client
pnpm install
pnpm dev

cd server
go mod tidy
go run ./cmd/subvision/main.go

cd vfx
pnpm install
pnpm dev

/client   # Next.js frontend
/server   # Go backend
/vfx      # Node.js Remotion-based subtitle/effects renderer

PORT=8000
TMP_DIR=/tmp

CLIENT_URL=http://localhost:3000

MINIO_HOST=minio
MINIO_PORT=9000
MINIO_ACCESS_KEY=minio
MINIO_SECRET_KEY=minio123
MINIO_BUCKET=subvision

RABBITMQ_HOST=rabbitmq
RABBITMQ_PORT=5672
RABBITMQ_USER=guest
RABBITMQ_PASSWORD=guest

# Default to third_party/whisper.cpp/bindings/go/models/*.bin
WHISPER_MODEL_PATH=third_party/whisper.cpp/bindings/go/models/ggml-base.en.bin

NEXT_PUBLIC_SERVER_URL=http://localhost:8000

TMP_DIR=tmp

MINIO_HOST=minio
MINIO_PORT=9000
MINIO_ACCESS_KEY=minio
MINIO_SECRET_KEY=minio123
MINIO_BUCKET=subvision

RABBITMQ_HOST=rabbitmq
RABBITMQ_PORT=5672
RABBITMQ_USER=guest
RABBITMQ_PASSWORD=guest

1. Integrate with Remotion for advanced video rendering.
2. Add parameter requests from client for user subtitle preferences:

• Language
• Color
• Outline
• Border
• Font name
• Font size
• Vertical margin (vmargin)

3. Add functionality to split subtitle segments by max characters or by seconds.
4. Add configuration for users to resize the video.

MIT

Made with ❤️ for creators.