jonesckevin/live-translate repository overview

⁠Live Translate

A self-hosted real-time translation web application with speech-to-text, virtual keyboards, and dual-panel conversation mode. Supports LibreTranslate (offline) and 10+ LLM providers as long as you bring your own Key (BYOK). The key can be set at the user or server level.

⁠Features

• Text Translation – Default uses LibreTranslate Framework to Translate between 37+ languages with auto-detection
• Dual Engine – LibreTranslate (Argos, self-hosted, offline) + Optional LLM providers (OpenAI, Anthropic, Gemini, Ollama, LM Studio, DeepSeek, Cohere, Groq, Grok, Mistral, Perplexity)
• Live Conversation Mode – Dual-panel real-time translation with per-panel language and microphone selection
• Speech-to-Text – Web Speech API (browser-native) + optional Whisper (server-side via faster-whisper) + optional AI provider STT (for supported providers/models)
• Custom Glossaries – Create term glossaries for consistent translations
• Session Management – Auto-save conversation history with export
• Docker Deployment – Single docker compose up

⁠Quick Start

⁠Portainer

1. Link to Github repository: github.com/4n6post/live-translate
2. Load .env file into Portainer stack environment variables
3. Modify Any Variable you want to change (optional)
4. Deploy the stack and open http://localhost:9015⁠ in your browser.

⁠Local Build with Docker Compose

git clone https://github.com/4n6post/live-translate.git
cd live-translate

# Start the application with automatic image building
docker compose up -d --build

During image build, Whisper models tiny, base, small, medium, and large-v3 are pre-downloaded into an internal seed cache, then copied into /data/whisper-model on first container start. This avoids runtime re-downloads and persists models across restarts/rebuilds.

Embedded LibreTranslate language package downloads are also persisted by mounting ./data/cache/libretranslate-local to /root/.local, so Argos language models are reused across container rebuilds/restarts.

Open http://localhost:9015⁠ in your browser.

⁠Quick Start

⁠DockerHub Run

docker run --name live-translate --restart unless-stopped -p 9015:5000 -v ./data:/data \
	-e ALLOW_AUTH=true \
	-e REQUIRE_AUTH=true \
	-e SOCKETIO_CORS_CREDENTIALS=true \
	-e ALLOW_USER_REGISTRATION=true \
	-e ALLOW_GUEST_LOGIN=true \
	-e WHISPER_ENABLED=true \
	-e WHISPER_MODEL=tiny \
	-e ENABLE_SERVER_ANALYTICS=false \
	-e SECRET_KEY='$(openssl rand -hex 32)' \
	-e SECRETS='$(openssl rand -hex 32)' \
	-e REQUIRE_SECRETS=true \
	-e LOGS_ACCESS_TOKEN= \
	-e ALLOW_CLIENT_API_KEYS=true jonesckevin/live-translate:latest

⁠Build

git clone https://github.com/jonesckevin/live-translate.git
cd live-translate

# Copy the example environment file and configure your settings
cp .env.example .env

# Start the application
docker compose up -d --build

The Dockerfile accepts PYTHON_BASE (default python:3.11-slim).

During image build, Whisper models tiny, base, small, medium, and large-v3 are pre-downloaded into an internal seed cache, then copied into /data/whisper-model on first container start. This avoids runtime re-downloads and persists models across restarts/rebuilds via ./data/whisper-model.

Embedded LibreTranslate language package downloads are also persisted by mounting ./data/cache/libretranslate-local to /root/.local, so Argos language models are reused across container rebuilds/restarts.

Open http://localhost:9015⁠ in your browser.

⁠Environment Variables

All environment variables are configured in the .env file. Copy .env.example to .env and configure as needed.