Bark Small

Bark Small TTS

A distilled 150M version of Suno's Bark that trades some fidelity for roughly twice the speed and half the memory.

Bark Small is Suno's distilled, lightweight take on Bark, shrinking the model to 150M parameters. It keeps Bark's signature ability to produce emotional speech, laughter, and multilingual output across the same 13+ languages, but generates noticeably faster (~8s versus ~15s) and fits in about 2GB of VRAM. The cost is some audio quality, so it works best as the go-to when full Bark is too slow for a creative task but you still want its expressive, sound-rich character. Like full Bark it caps at 200 characters per generation and does not support voice cloning.

At a glance

Developer
Suno
License
MIT
Tier
standard
Speed
medium
Voice cloning
No
Languages
English, Chinese, French, German, Hindi, Italian, Japanese, Korean, Polish, Portuguese, Russian, Spanish, Turkish
Max characters
200

Bark Small AI Voices

Chinese Speaker

Chinese
표준 Neutral
사용

English Female 1

English
표준 Female
사용

English Female 2

English
표준 Female
사용

English Male 1

English
표준 Male
사용

French Speaker

French
표준 Neutral
사용

German Speaker

German
표준 Neutral
사용

Hindi Speaker

Hindi
표준 Neutral
사용

Italian Speaker

Italian
표준 Neutral
사용

Japanese Speaker

Japanese
표준 Neutral
사용

Korean Speaker

Korean
표준 Neutral
사용

Polish Speaker

Polish
표준 Neutral
사용

Portuguese Speaker

Portuguese
표준 Neutral
사용

Russian Speaker

Russian
표준 Neutral
사용

Spanish Speaker

Spanish
표준 Neutral
사용

Turkish Speaker

Turkish
표준 Neutral
사용

Best for

Quick creative audio when full Bark is too slow

Bark Small TTS — FAQ

Bark Small is a distilled 150M-parameter version (versus 350M for full Bark). It runs faster — around 8 seconds per clip in ~2GB VRAM — at the cost of some audio quality, while keeping the same emotional and multilingual abilities.

Yes. Like full Bark, Bark Small is MIT-licensed and free for commercial use.

Choose Bark Small for quick, creative, emotive snippets when full Bark is too slow. Both share a 200-character cap, the same 13+ languages, and neither supports voice cloning.
← All voices