I spent weeks fighting Qwen3-TTS to get decent French voice cloning on my RTX 3090. Seed pinning, audio trimming, per-language prompts - nothing worked reliably. Then I found Chatterbox Multilingual, and everything just clicked.
A step-by-step guide to setting up Qwen3-TTS with GPU acceleration on Ubuntu Server. Complete with FastAPI wrapper, voice cloning via x-vector embeddings, and all the troubleshooting tips I wish I had when I started.
Deploy a 1.7B parameter TTS model with interesing featyures. No cloud APIs, no surprise bills, total data privacy.
Setting up Fish Speech text-to-speech on TrueNAS SCALE with Docker and NVIDIA GPU support. Covers the gotchas I hit along the way.
Exploring further the TTS options with an AI, the result is mesmerising
Text to speech with an AI, what can go wrong ?