Chinese Student Translates New Murakami Novel With AI in Four Hours

Four hours after Haruki Murakami's latest novel was released in Japan, a Chinese translation was already available for download online. It wasn't the work of a professional translator, but of a literature student who wrote a program to feed the book's text, chunk by chunk, into the DeepSeek language model. The entire operation cost less than $2.

A Program Instead of a Translator

Ping Ming, a Murakami fan who doesn't speak Japanese, had long wanted to read the author's new novels as soon as they came out. He had previously tried the free version of DeepL, but found the results unacceptable due to grammatical errors and awkward phrasing. Two years ago he taught himself to code in Python, and that skill became the key to solving the problem.

The tool he built, called Wenyi, splits the book's text into chunks of fewer than 2,000 words and feeds them one by one to the Chinese DeepSeek model with a request to translate. It translated Murakami's newest novel, "The Tale of KAHO," about a 26-year-old children's book author caught up in a series of strange events, in its entirety before any human had a chance to read the result.

Errors, Fixes, and a Quick Takedown

After publishing the text, Ping Ming noticed flaws, including inconsistent translation of the various nicknames used for the main character, Kaho, and passages with an artificial, machine-like style in places. Together with a collaborator using the handle Nanshan Youlu, he refined the code, improved how names were handled, and had the model smooth out the text further.

The translation spread online, including to the Z-Library archive, Chinese social media platforms, and the resale site Xianyu. The first post linking to the translation gathered over 2,000 likes, with users praising the quality of the text. Shortly after finishing the final chapter, the creators took their own version down, explaining that Wenyi's translation had turned out to be too faithful, rendering the original Japanese text almost word for word.

Professional Translators Weigh In

The case has divided China's literary community. Lin Shaohua, a veteran translator of earlier Murakami novels including "Kafka on the Shore" and "Norwegian Wood," told the magazine China Newsweek that Wenyi's text reads smoothly but fails to capture the author's voice.

AI translation has reached a level that's grammatically coherent and reads fluently, but that's about it. - Lin Shaohua, translator of Murakami's novels

AI still can't capture subtle linguistic and stylistic nuances. - Lin Shaohua, translator of Murakami's novels

A Copyright Problem

Both Ping Ming and his collaborator risk copyright infringement claims for distributing the translation without a license, regardless of the fact that they made no financial profit from it. Murakami's novels typically reach official translations in other languages with significant delays, which has for years fueled a market for unauthorized, fast fan translations across Asia.

A similar pattern has emerged outside China as well. In Iran, at least six Persian versions of "The Tale of KAHO" have been produced or are being prepared in parallel, some based on an unofficial English text bearing signs of AI generation, meaning some translations may rely on an intermediate AI-generated text rather than the authorized original.

What It Means for the Translation Industry

Ping Ming's story shows just how much shorter the path from a book's release to its unofficial translation has become, now that a language model can replace months of a translator's work with a few hours and a cost comparable to a cup of coffee. It's another sign that models like DeepSeek handle long, coherent literary prose far better than earlier tools like DeepL, the very tool Ping Ming had complained about before.

For publishers and translators, this creates pressure from two directions: readers expect access to bestsellers immediately after release, and machine translation quality has improved enough that a student's amateur project can compete with professional work, at least in terms of textual fidelity. Whether it also captures the author's literary voice remains an open question, one that divides translators themselves.