Startup claims to beat humans, ChatGPT in literary translations

This article is a part of Startup Spotlight, a series that features young, up-and-coming startups.

Image credit: Shutterstock

Tales are being left untold, locked behind linguistic walls. These stories end up with limited reach due to high translation costs.

Omni, an Indonesian startup specializing in literary translations, is addressing this challenge with an AI-powered approach that it claims outperforms humans and even off-the-shelf products from tech giants.

😟 Problem

Despite the growing demand for multilingual content, the complex nature of literary texts makes them hard to translate.

Machine translations, even with the advent of large language models (LLMs), are still far from viable for high-quality literary work, the startup says.

This is due to a few factors:

• Literary translation requires understanding rich language and cultural nuances, which often gets lost in machine translation.
• Current translation methods face limited training data.
• Literary works demand long-range contextual understanding, which isn’t fully addressed even by LLMs with long context windows.

💡 Solution

Omni claims to outperform both human translators and AI tools like ChatGPT and Google Translate. It uses the following methods:

• Custom-trained LLMs for better literary style and accuracy.
• An agentic translation method
• A proprietary algorithm for ensuring translation coherence of long texts
• An ability to collect training data through its consumer-facing translation app.

In the future, it plans to launch an enterprise product that will support more languages, even better translations, as well as image and audio.

📊 Market Size

The global translation services market is valued at about US$42.2 billion. Literary translation, including those of webnovels, forming about 10% of this market—an estimated US$4.2 billion.

🤝 Team

🛞 Pilot project

🏆 Competition

💰 Financials