[Plex] Auto-dub movies in multiple languages using AI

Hey everyone! 👋 I built a tool that automatically dubs movies in your Plex library into other languages using open-source AI — with voice cloning that keeps the original actor's voice. ## What it does You tag a movie in Plex with a label like `dub-spanish`, and a background service picks it up, runs it through an AI pipeline, and adds a new audio track to the file. Next time you play the movie in Plex, you can select the dubbed language from the audio menu — just like Netflix. ## The AI pipeline 1. **Demucs** — separates vocals from music/sound effects 2. **Whisper** — transcribes the dialog 3. **Meta NLLB-200** — translates to the target language 4. **Coqui XTTS v2** — clones the original actor's voice and speaks the translation 5. **FFmpeg/MKVToolNix** — muxes the new audio track back into the file ## How it runs The installer creates a dedicated Proxmox LXC with: - GPU passthrough (NVIDIA) for acceleration - NAS mounts auto-detected from your existing Plex LXC - All dependencies pre-installed (Python, PyTorch, Open Dubbing) - A `dubber` CLI command and optional label-watcher systemd service The setup script auto-detects your GPU, driver version, storage, templates, and media mounts. It bind-mounts the host's NVIDIA libraries into the LXC to avoid driver version mismatches. ## Requirements - Proxmox VE 7.x or 8.x - NVIDIA GPU (12+ GB VRAM recommended for GPU mode, CPU fallback for smaller GPUs) - Media on NFS/CIFS share or local storage - Free HuggingFace account (for speaker diarization models) ## Supported languages (voice cloning) English, Spanish, French, German, Italian, Portuguese, Polish, Turkish, Russian, Dutch, Czech, Arabic, Chinese, Hungarian, Korean, Japanese, Hindi Edge TTS (no voice cloning) adds 60+ more languages including Hebrew, Swedish, etc. ## Quick start ```bash git clone : Link URL cd Dubber_LXC bash setup_dubber_lxc.sh ``` The installer walks you through everything interactively. ## Known limitations - Voice cloning quality is decent but not perfect (open-source models are improving fast) - GTX 1060 (6GB) works but must run on CPU — 12+ GB VRAM recommended for GPU mode - Hebrew voice cloning not yet supported by XTTS v2 (Edge TTS works as fallback) - First run downloads ~5GB of AI models This is my first open-source project — feedback and contributions welcome! 🙏