BlueRoseLabs releases Silencio-2 with emotion tags and voice cloning
Silencio-2 is a Kyutai Pocket TTS 18M fine-tune for English speech. It supports emotion and sound tags and reference-based voice cloning and streams 24 kHz mono audio. The model card lists 27.9M total parameters, with 18.2M used for speech generation, and licenses the weights under CC BY 4.0 and core inference code under MIT.
Visit original source