Bark: Transformer-based Text-to-Audio Generation Model by Suno-ai

Bark is a transformer-based text-to-audio model created by Suno. It can generate highly realistic, multilingual speech as well as other types of audio, including music and simple sound effects. The model also has the capability to produce nonverbal communications like laughing, sighing, and crying. Bark supports various languages out-of-the-box, automatically determining the language from input text. It also has voice presets and voice/audio cloning, the capability to fully clone voices including tone, pitch, emotion, and prosody. Bark uses GPT-style models to generate audio from scratch and can generalize to arbitrary instructions beyond speech that occur in the training data. It is available for installation via pip and can be run on both CPU and GPU. Bark is licensed under a non-commercial license, CC-BY 4.0 NC, and EnCodec, which functions as an audio codec, is licensed under a non-commercial license.