Sakana AI has added a brand new function known as Sakana Translate to its chat service, Sakana Chat. It handles bidirectional translation throughout Japanese, English, and Chinese language. The interpretation engine is Namazu, the corporate’s mannequin sequence tailored for Japanese.
Sakana Translate ships as a free internet app. A single account unlocks all three of its modes.
What’s Sakana Translate
Sakana Translate is a browser-based translation product, not a brand new base mannequin. It runs on Namazu, Sakana AI’s Japan-adapted mannequin sequence.
The idea Sakana AI states is ‘deep translation for Japan.’ The aim goes past swapping phrases and sentence constructions. It goals to hold context, tone, and register between languages.
That focus targets a particular hole. Sakana AI argues that basic instruments usually miss what makes Japanese distinctive. Examples embody enterprise honorifics, culturally particular ideas, abbreviations, and web slang. Grammar could keep appropriate whereas interpersonal tone will get misplaced.
The product bundles three features into one display screen: Translate, Proofread, and Ask.
The Three Modes, Defined
Every mode targets a unique on a regular basis process. The desk beneath summarizes them.
| Mode | What it does | Key element | Greatest for |
|---|---|---|---|
| Translate | Converts pasted textual content between the three languages | As much as ~5,000 Japanese characters, streaming output, historical past saved robotically | Emails, slide decks, articles, internet pages |
| Proofread | Refines a draft right into a extra pure model | Adjustments proven with diff highlighting; adjusts tone, politeness, and ritual | Enterprise e-mail and English writing checks |
| Ask | Solutions follow-up questions on a end result | Clarifies nuance, suggests alternate options, explains grammar in the identical context | Studying why a translation reads the best way it does |
A number of phrases are value unpacking for engineers new to this house.
- Streaming output means the interpretation seems progressively, token by token. You don’t look ahead to the complete end result earlier than studying. This mirrors how chat fashions return textual content.
- Diff highlighting exhibits precisely what modified. Additions and removals are marked inline, like a version-control diff. Proofread goes previous grammar. It additionally tunes naturalness, politeness, and the register a reader expects.
- Ask removes the tool-switching drawback. You now not bounce between a translator and a dictionary. Nuance questions get answered in opposition to the identical supply and output.
You possibly can strive all three modes within the interactive demo beneath
