Independent testing from Japan
We test transcription tools with real Japanese audio, then a native Japanese reviewer checks every number, name, and kanji choice by hand.
Updated August 9, 2026 · Four controlled recordings
CONTEXT TEST
橋の端で箸を落とした。
Preliminary—not a final winner. More real-world tests are coming.
01 Recorded in Japan
02 Reviewed by a native speaker
03 Same audio for every tool
04 Weaknesses published
Early benchmark results
These are hands-on results from four short controlled recordings. Test 04 currently includes Notta and Sonix; TurboScribe and Otter are still pending.
The buttons below are direct product links. We do not earn a commission from them yet. Any future affiliate link will be clearly labeled beside the button.
MEETING & NOTES
No major meaning errors across four recordings—and it separated all nine turns in our first two-speaker test.
Watch out: Numbers were usually written in kanji, and the invented name ミライテック became 未来テック.
AUDIO TRANSCRIPTION
Produced the cleanest casual-speech result. Test 04 is pending because the daily free allowance was exhausted.
Watch out: 橋の端で箸 became 橋の橋で橋, despite the polished-looking output.
AUDIO TRANSCRIPTION
Matched Notta with all nine speaker turns correctly labeled and no content errors in the first two-speaker test.
Watch out: It wrote 会議室B as 会議室日 in test 03 and 橋の端で箸 as 橋の橋で橋 in test 02.
MEETING & NOTES
Recognized ミライテック and two of three homophones, but changed a price. Later tests were blocked by the Basic plan's import cap.
Watch out: 1万2,800円 became 12,008百円—a materially different amount if read literally.
WEB TRANSCRIPTION
The no-signup web converter accepted all four recordings but failed during processing every time.
Watch out: This is a web-tool reliability finding, not an accuracy verdict. The signed-in app may behave differently.
Biggest lesson so far
Polished text can still be confidently wrong.
Sonix reported high confidence while writing 橋の端で箸 as 橋の橋で橋, then heard 会議室B as 会議室日. Otter changed ¥12,800 into a materially different amount, while JotMe failed to return text for four different files. Native review still matters.
Our method
Every tool receives the same original audio. We separate recognition mistakes from punctuation and formatting, and we do not score a tool that fails to produce a transcript.
A short, naturally paced recording used to check basic content accuracy and punctuation.
Dates, prices, an address, an invented company name, ChatGPT, and 橋・端・箸 in context.
Fillers, four self-corrections, shortened phrases, times, responsibility, and agenda details.
Quiet-room result: Notta and Sonix both labeled 9/9 speaker turns correctly. TurboScribe and Otter are pending because their free import limits were reached.
A transcript can look polished to a non-Japanese speaker while getting names, kanji, politeness, or meaning wrong. Native Japanese AI Lab is built in Japan to make those mistakes visible—and help international users choose with confidence.
The product buttons currently use ordinary direct links, so we earn no commission. If affiliate links are added later, each one will be labeled clearly beside the button. Compensation never changes our test files, scoring method, ranking, or the weaknesses we publish.