Suno open-source Transformer-based audio generation model capable of expressive speech, music, and ambient noise.
Pretrained checkpoints and source code fully open-sourced by Suno under MIT License for unrestricted use
Bark is an influential open-source audio generation model developed and released by Suno under the MIT license. Unlike traditional phoneme-based speech synthesizers, Bark leverages a GPT-style Transformer to generate discrete EnCodec acoustic tokens straight from text prompts. It is celebrated for expressive non-speech sounds—including laughter ([laughs]), sighs ([sighs]), throat-clearing, whispering, and ambient soundscapes. With built-in multilingual speaker libraries, Bark remains a foundational landmark for expressive generative audio research.