Spark TTS

What is Spark TTS?

Spark TTS is an advanced text-to-speech system that leverages the powerful capabilities of large language models (LLM) to achieve highly accurate and natural speech synthesis. This system stands out with its zero-shot voice cloning and fine-grained voice control features, representing a major breakthrough in the field of speech synthesis. Spark TTS is efficient, flexible, and powerful, suitable for both research and production use.

Spark TTS Key Features

1. Concise and efficient:

Spark TTS is entirely built on Qwen2.5 and does not require additional generative models like stream matching. It reconstructs audio directly from the codes predicted by the LLM without relying on separate models to generate acoustic features. This approach simplifies the process, improves efficiency, and reduces complexity.

2. High-quality voice cloning:

Spark TTS supports zero-shot voice cloning, meaning it can replicate a speaker’s voice even without specific training data for that voice. This is well-suited for cross-lingual and code-switching scenarios, allowing seamless transitions between languages and voices without separate training for each.

3. Bilingual support:

Spark TTS supports both Chinese and English, with zero-shot voice cloning capabilities in cross-lingual and code-switching scenarios, enabling the model to synthesize speech in multiple languages with high naturalness and accuracy.

4. Controllable speech generation:

Supports creating virtual speakers by adjusting parameters such as gender, pitch, and speech speed.