armenian-video-dubbing

Deployment Guide

Docker (Recommended)

Quick Start

# Build the image
make build

# Start all services
docker compose up -d

# Check status
docker compose ps

Service Architecture

Service	Port	Description
nginx	80, 443	Reverse proxy with rate limiting
gradio	7860	Web UI
api	8000	REST API
label-studio	8080	Annotation tool (dev profile)

Environment Variables

Copy .env.example to .env and configure:

cp .env.example .env

Required:

• HF_TOKEN — HuggingFace access token for model downloads

Optional:

• ARMTTS_API_KEY — Enable API authentication
• CUDA_VISIBLE_DEVICES — GPU selection
• WANDB_API_KEY — Experiment tracking

GPU Access

Docker Compose is configured with NVIDIA GPU support. Ensure you have:

• NVIDIA Container Toolkit installed
• Docker configured for GPU access

# Verify GPU access
docker run --rm --gpus all nvidia/cuda:12.4.0-base-ubuntu22.04 nvidia-smi

SSL/TLS

Place your certificates in docker/ssl/:

docker/ssl/
├── cert.pem
└── key.pem

The nginx configuration will automatically use them.

---

Cloud Deployment

RunPod

# Set your API key
export RUNPOD_API_KEY=your-key

# Deploy
make deploy-runpod
# or
bash scripts/deployment/deploy_runpod.sh

AWS (EC2 + GPU)

# Configure in .env
AWS_REGION=us-east-1
AWS_INSTANCE_TYPE=g5.xlarge
AWS_KEY_NAME=armtts-key

# Deploy
bash scripts/deployment/deploy_cloud.sh aws

GCP (Compute Engine + GPU)

# Configure in .env
GCP_ZONE=us-central1-a
GCP_MACHINE_TYPE=n1-standard-8
GCP_GPU_TYPE=nvidia-tesla-t4

# Deploy
bash scripts/deployment/deploy_cloud.sh gcp

Cost Estimation

python scripts/deployment/cost_estimate.py

Estimates per-minute dubbing costs across 6 cloud providers.

---

Manual Deployment

System Requirements

• Ubuntu 22.04+ / macOS 13+
• Python 3.11+
• NVIDIA GPU with CUDA 12.4+
• FFmpeg, rubberband-cli, libsndfile

Install

# System deps (Ubuntu)
sudo apt-get install -y ffmpeg libsndfile1 rubberband-cli

# Python environment
bash scripts/setup_environment.sh

# Verify
python scripts/verify_setup.py

Run Services

# Gradio UI
make web

# FastAPI
make api

# Both with nginx (requires nginx installed)
# Configure docker/nginx.conf for your setup

---

Monitoring

Prometheus Metrics

The API exposes metrics at /metrics:

• armtts_requests_total — Total API requests
• armtts_dub_duration_seconds — Dubbing job duration
• armtts_active_jobs — Currently processing jobs
• armtts_gpu_memory_used_bytes — GPU memory usage

Health Checks

# API health
curl http://localhost:8000/api/v1/health

# Gradio health
curl http://localhost:7860/

Docker Compose includes health checks with 30s intervals.

This site is open source. Improve this page.