AI Paper Translator

Translate, summarize, and extract from research papers in seconds.

Output language

Choose your preferred summary language

Drop your PDF here

PDF files up to 10MB · Supports research papers in any language

How it works

From PDF to a translated and citable summary

One upload triggers a full pipeline: translation, summary, and grounded follow-up Q&A. All in your language.

Step 1

Upload PDF or paste DOI / URL

Drop a PDF, paste a DOI, or paste a paper URL. No account needed for your first paper.

Attention Is All You Need

2017Research Paper1,247 citations

A. Vaswani, N. Shazeer, N. Parmar et al.

EnglishOriginal

日本語

Abstract

The dominant sequence transduction models are based on complex recurrent or convolutional neural networks that include an encoder and a decoder. We propose a new simple architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely.

Figure 1. Transformer — model architecture.

Key Findings

Self-attention layers replace recurrence for sequence modeling
Multi-head attention integrates information from different subspaces
Achieves 28.4 BLEU on WMT 2014 English-to-German translation

Step 2

Get translation + summary in your language

Field-aware translation preserves formulas, citations, and figure references. Auto-summary in EN or JA delivered in seconds.

Attention Is All You Need

2017Research Paper1,247

A. Vaswani, N. Shazeer et al.

Self-attention is a mechanism relating different positions of a single sequence to compute a representation. The Transformer is the first transduction model relying entirely on self-attention.

The Transformer allows for significantly more parallelization and can be trained in as little as twelve hours on eight P100 GPUs — a fraction of the time required by recurrent models.

On the WMT 2014 English-to-German translation task, the big Transformer model establishes a new single-model state-of-the-art BLEU score of 28.4.

Kenkyu AI

The Transformer models long-range dependencies with self-attention1.2, trains much faster than recurrent architectures1.4, and reaches 28.4 BLEU on WMT 2014 English–German1.7.

Ask a follow-up question...

Step 3

Ask up to 3 grounded follow-up questions

Every answer cites the exact passage from the source paper, so you can verify accuracy before quoting it.

WANT JUST A SUMMARY?Get an instant, citation-linked summary of any paper with Paper Summarizer.

Open Paper Summarizer

Privacy & your papers

Your unpublished work stays yours

Upload confidential drafts, peer-review manuscripts, and internal lab papers without compromise.

No trainingAnonymousEncrypted

Never used for model training

Your uploads are not added to any training set. We don't train models on user-submitted papers, period.

Uploaded files deleted after processing

Your uploaded PDF is removed from storage as soon as text extraction finishes. Your anonymous session itself expires after 24 hours.

No account required for one paper

Sensitive uploads can skip registration entirely. Anonymous sessions get a full translation, summary, and 3 follow-ups.

Encrypted in transit

All traffic is protected with TLS, and uploads are kept in an isolated anonymous storage environment. AI processing runs on API-tier endpoints that are never used to train models.

Why Kenkyu.ai

How it compares to DeepL, Elicit, ChatGPT, and others

A side-by-side view of what each tool actually delivers when you upload a full research paper.

Tool

Japanese output

Full-paper translation

Q&A with citations

Kenkyu.aiPaper Translator

DeepLGeneral translator

Google TranslateGeneral translator

ElicitResearch assistant

SciSpaceResearch assistant

ChatPDFChat-with-PDF

ChatGPTGeneral AI assistant

Built for non-native English researchers

Cut time spent on literature reviews in half

Non-native readers spend nearly double the time on English papers as native speakers. Kenkyu.ai closes that gap so you can read, verify, and cite at the speed of your peers.

Graduate students

Reading a foreign methodology paper for a thesis section

You need to understand a method paper line-by-line, but English jargon plus unfamiliar formulas means hours of re-reading.

Read in Japanese, verify the formula against the source, and ask the paper itself to clarify the missing assumption.

Postdocs

Preparing a literature review where 70% of sources are in English

Skim-reading 40+ papers in a second language; quoting any of them risks subtle mistranslation.

Summary first, full translation when it matters, and grounded Q&A to confirm a quote before you commit to citing it.

Journal-club leads

Summarizing this week's paper for a JA-speaking lab

Re-writing an English abstract into accurate Japanese for ten colleagues is everyone's least favorite task.

Drop the PDF, share the Japanese summary, and use the 3 follow-ups to nail the questions your team will actually ask.

Industry researchers

Screening papers outside your language or field

Most papers won't be worth deep reading, but you can't decide without understanding them first.

Triage at translation speed: scan the summary in your language, dig into the full paper only when it looks relevant.

Pricing

Accelerate Your Research

Try out Kenkyu.ai completely for free, and upgrade to Plus for unlimited chats, file uploads, and more.

Save 40%

Free Plan

¥0/month

Get to know how Kenkyu.ai can help you, stress free.

Unlimited search across 230 million papers
AI Chats: 10
File uploads: 10
File / Translation size limit: 25mb
File / Translation character limit: 200,000 (~100 pages)
Email support within 2 business days

Plus Plan

¥1,800/month

Billed yearly at ¥21,600/year

Unlock unlimited research.

Unlimited search across 230 million papers
AI Chats: Unlimited
File uploads: Unlimited
File / Translation size limit: 100mb
File / Translation character limit: 1,000,000 (~500 pages)
Email support within 2 business days

Enterprise Plan

Custom

For those who do research at scale.

Everything on Plus Plan
Prioritized support
And more…

FAQ

Questions researchers actually ask

Is the translation accurate enough to cite in my paper?

It's accurate enough to read confidently, but the responsible workflow is to verify any specific quote against the source PDF. Every Q&A answer cites the exact passage so you can do that in one click.

Which file formats does it accept?

File upload accepts PDF only. For DOI links, PDFs are extracted where available. For URLs we capture the content on the webpage.

How long can the paper be?

We accept PDFs up to 10MB, and apply a 100,000 character limit to both PDFs and URLs.

Which languages are supported for the summary?

English and Japanese, in both directions. Translation between EN and JA is fully supported; additional language pairs are rolling out.

Can I translate from Japanese to English?

Yes, and the same Q&A pipeline works in reverse. Useful for JA researchers preparing English-language submissions or sharing JA papers with overseas collaborators.

Is my uploaded paper private?

Yes. Your uploaded PDF is processed only under your session, is removed from storage as soon as it's processed, and is never used to train models. Your anonymous session expires after 24 hours.

Why only 3 questions on the free tier? What unlocks more?

3 questions cover the typical 'is this paper worth deep-reading?' check. Creating a free account increases the limit, and our Plus plan is completely unlimited.

How is this different from DeepL or Google Translate?

DeepL and Google Translate can smoothly translate a limited amount of text, but become very inconvenient for full papers. With Kenkyu.ai you additionally have the ability to save your uploads into a library and ask citation-grounded questions about your papers.

How is this different from Elicit or SciSpace?

Elicit and SciSpace are research assistants that don't ship a native Japanese interface or full-paper translation. Kenkyu.ai is JA-first, returns a complete translation (not just an answer), and cites each Q&A passage back to your uploaded source.

Is there a usage limit per day?

Anonymous sessions allow one paper per 24-hour window. Creating a free account raises this limit, and our Plus plan removes it. Rate limits exist to protect our service, not to push upgrades.

Translate your next paper now.

Free and anonymous. No credit card and no email for your first paper. Keep going with a free account.