Bark

What is Bark?

Bark is an open-weight text-to-audio model released by Suno AI and published on GitHub under an MIT licence. It speaks text, but it also emits things a conventional text-to-speech engine will not: laughter, sighs, gasps, breathing and fragments of background noise, which suits game dialogue and audio sketches. Output follows a voice preset taken from short reference clips, so timbre can be steered without training anything. It runs locally on a GPU, so there are no per-character fees and no audio leaving your machine. The differentiator is that this is a research model you host yourself, not a hosted speech service.

Last updated: 2026-09-20. This site only provides an index; for exact features, pricing, and licensing, see the official website.

Key features

• Text-to-speech plus laughter, sighs, gasps and other non-speech sounds
• Multilingual output covering around a dozen major languages
• Voice presets triggered by short reference audio clips
• Runs locally on a CUDA GPU with released model weights
• Python library plus a simple command line generation script
• MIT licence permits commercial use and modification of the code

Pros & cons

Strengths

• Free to run locally with no per-character or per-minute fees
• Produces laughter and sound effects that plain TTS engines skip
• Open weights can be inspected, fine-tuned and self-hosted

Watch out for

• Slow generation and heavy GPU memory use next to hosted APIs
• Long passages drift, hallucinate or lose the speaker identity
• No support, hosted endpoint or commercial SLA behind the project

Best for & use cases

game dialogue prototypes, audio experiments, sound design and offline speech generation

If you're comparing similar products, check the alternatives below, or browse all tools in the AI Audio & Music category.

FAQ

Similar tools