A 3 billion parameter German fine-tuned text-to-speech model from Canopy Labs, optimized for research applications.