Skip to content

Hibiki

A streaming model for speech translation

Visit official site

What Hibiki is

The model is described as a streaming speech translation model that generates correct translation chunk by chunk as context builds.

What you can do with it

  • Generate target speech and text translation
  • Run a chosen backend on your hardware

Pros & cons

✓ What we like

  • Low-latency real-time translation is listed
  • Multiple backends such as PyTorch, Rust and MLX

! What to watch out for

  • The named language pair is French to English
  • Translation is built chunk by chunk on context

FAQ

Which languages are supported?

The material names French to English translation.

Is it open?

The model is described as open and free for developers and researchers.

Last reviewed: 2026-11-15

More AI audio tools tools

View all →

How we review