Bark TTS

Suno's transformer-based text-to-audio model that generates speech plus laughter, sighs, music, and sound effects.

Bark comes from Suno and takes a different approach from most TTS systems: it is a GPT-style transformer trained as a text-to-audio model rather than a pure text-to-speech one. Because it generates raw audio tokens, it can produce nonverbal sounds — laughing, sighing, crying — as well as background music and sound effects alongside the spoken words. It ships with 100+ speaker presets and handles 13+ languages including English, Chinese, French, German, Hindi, Japanese, and Korean. The trade-off is speed and length: at 350M parameters it runs slowly (~15s per clip) and caps at 200 characters, so it shines for short, emotive, creative audio rather than long narration.