Background Image

Oliver's Pronunciation Library

This is a library of Japanese pronunciations made to cover words, expressions and abbreviations that JapanesePod101 doesn't (such as 即, 伝家の宝刀, and ウーロン). This library is optimised for use with Anki, and can be hooked up to Yomichan.


This library was curated while using Yomichan and various dictionaries (most notably JMdict). It was initially just a project of mine to supplement pronunciations, but later I chose to create a service that will hopefully make other people's lives easier. While it is not comprehensive, this library aims to cover the majority of pronunciations that JapanesePod101 doesn't.

Audio files here were derived from Forvo. Having some knowledge in sound design, I've edited a lot of pronunciations to be more consistent and usable in Anki, making OPL advantageous in terms of quality.

Here are some comparisons between original and edited pronunciations sourced from Forvo.

期末テスト:

ゴミカス:

ドット絵:

Generally, all audio files will peak around -6db.

Additionally, this server has the advantage of accounting for any combination of kana in readings (for instance, the reading きまつテスト will match with きまつてすと). Generally, this will help cover words with varied kana readings.

While this library is curated by myself, I can accept additional pronunciation files provided they meet the basic sound consistency. Please contact me for this, and preferably include licences and sources if applicable.

Connecting via Yomichan is simple:

  1. Open Yomichan's settings and navigate to "Configure audio playback sources...".

    Image 1Image 1
  2. Add a new audio source and select Custom URL.

    Image 2Image 2
  3. Input https://opl.oliverspeak.com/search?term={term}&reading={reading} as the URL.

    Image 3Image 3

I suggest you prioritise OPL if you are also sourcing directly from Forvo.
It's worth noting that OPL will have a few pronunciations already covered by JapanesePod101, but are poor quality (too quiet, etc).

Right now, some caveats exist with the Yomichan integration.
For instance, some pronunciations may be incorrect when there are multiple readings for a word, or if there's variation in pitch accent. This only happens when there isn't an exact pronunciation, and the server chooses a file based on the term, rather than the reading.

Alternatively, use https://opl.oliverspeak.com/based-search?term={term}&reading={reading} for 100% coverage of all pronunciations guaranteed.
Currently serving 1079 files.

Return to oliverspeak.com

This page is licenced under the Apache License 2.0. The source code can be found here.

All pronunciations sourced from Forvo are subject to the Forvo Media S.L. license, which can be found here.