Show HN: I built a fast, accurate translation API for 1/30th the price of Google

Show HN: I built a fast, accurate translation API for 1/30th the price of Google(translateapi.web.app)

15 points by nbened 2 years ago | 14 comments

Hey HN,

I built an alternative to Google translate that's fast, accurate, and 1/30th the price. 20+ supported languages, not billed until 100k tokens, and already used thousands of times with real language learning users.

The landing page has a demo for you to test performance yourself for free with no card required. (Also landing page says 1/13 the price because had to account for all 3 big names, with varying prices each).

I built this as a workaround for my language learning startup, and it was perfect at a fraction of the cost. Hopefully it can help other devs in accessibility and language learning apps.

Thanks,

nbened 2 years ago |

Also, please hit me with any feedback, specifically for the documentation clarity and call quality.

My goal is to keep fundamental utilities and accessibility tools like translation services in the hands of small creators rather than some mega corporation, so I'm open to any improvement that can make that a reality.

Nic

android521 2 years ago | |

Any chance you can include Burmese language?

nbened 2 years ago | | |

I tried to, but the quality just isn't there. Eventually I'd like to, only if I can get the accuracy down.

lpellegr 2 years ago |

The idea is great but you will never compete with a solution like Google translate: one simple reason is that Google Translate is highly specialized and optimized. Did you try to compare the processing time?

Running an opensource AI model to perform translation may help and work in some cases but to be honest this has also a lot of limitations. The first being processing time and the second languages support. I am not even considering accuracy.

nbened 2 years ago | |

Thanks for the feedback, you bring some good points to the table.

Did you try the demo on the landing page and notice considerable latency/accuracy slip ups?

pierotofy 2 years ago |

Nice project!

A few questions/suggestions:

- It's not clear what the character limit would be if one was to sign-up. - What does "context-driven translations for higher accuracy" mean? - 35 characters are quite short for testing the translation of a complete sentence. Is there a reason for using such a short character length?

nbened 2 years ago | |

Hey, thanks for the feedback!

The character limit for an api call is 3000. I put it in the docs but could probably make it more obvious.

The context driven for higher accuracy is referencing what's going on under the hood, but on second thought it doesn't make much sense to put that there if I'm not going in depth about the under the hood stuff. I'll take that out.

35 character is quite short, you're right. I'll double it. The reason it was so short was arbitrary. I made it 100 now.

If you've got anything else send it over !

quickthrower2 2 years ago |

Nice. Do you have a secret sauce to get the price low or is it just a case of the usual overcharging by cloud providers?

Also read your twitter bio: I wont repeat here for privacy but would enjoy seeing stories about those things posted on HN

nbened 2 years ago | |

To the first question: A mix of both really. When you have a service this widespread, like translations, there's a lot in the vertical integration that can be more efficient. I mean, the project initiated from a different startup I built as a workaround for the cloud providers. It just evolved from that bottom-up thinking.

Thanks! Just for clarity, are you talking about the work experience, living situation, or startup ventures?

quickthrower2 2 years ago | | |

All of the above!

arcastroe 2 years ago |

Tha translations worked great in my testing between Spanish/English.

How does it work? Is it essentially a fine-tuned LLM? If so, which base model?

sanjogg 2 years ago |

I gave it a shot! Works really well.

nbened 2 years ago | |

Awesome!