Listen to any web article — paste a URL into Yomite's Web Reader

Your “read later” pile keeps growing because reading takes a dedicated hour you don’t have. Yomite’s Web Reader turns those waiting minutes into reading minutes — paste a URL, get a clean spoken article with no ads, no nav, no comments.

Three steps

1. Copy a URL

From your browser, a Twitter thread, a Slack link, an email. Anything.

2. Open Yomite

Tap the “Web リーダー” (Web Reader) tab at the bottom.

3. Paste and “読み込み” (Load)

Paste the URL → Load. Yomite fetches the article and runs 記事抽出 (the same library Firefox and Pocket use) to extract the body text.

What’s read aloud: title, author, body. What’s not: ads, navigation, related-articles widgets, comment threads.

Sites we’ve verified

記事抽出 works on a long list of sites. Tested and reliable:

News

• NHK NEWS WEB
• Nikkei Online (free articles)
• Asahi Shimbun Digital (free articles)
• Mainichi Shimbun (free articles)
• Yomiuri Online
• Yahoo News
• Bloomberg Japan
• Reuters JP
• ITmedia / @IT
• CNET Japan / TechCrunch Japan

Blogs and long-form

• note
• Hatena Blog & individual Hatena bookmark articles
• Qiita / Zenn
• WIRED Japan
• Huffington Post

English sources

The voices are Japanese, so English text gets read with a Japanese phonetic interpretation — fine for skimming, weird for serious listening. But it works on:

• Medium
• The New York Times
• The Atlantic
• BBC News
• The Verge

Paywalls and login pages are out of scope

Yomite can’t bypass paywalls or login walls. That includes:

• Nikkei paid members-only articles
• Asahi members-only articles
• NYT articles past the free quota
• Sites that require an account to view
• Some SPAs that render body text purely via JavaScript

Workaround: get the article into readable form in your browser first (or just copy-paste the text), then drop it into Yomite’s “Text Reader” feature.

Build a “to listen” queue

Tap to save any article to your Library. Saved articles:

• Work offline (the extracted body is cached locally)
• Track read / unread state
• Can be sorted into folders
• Support batch operations (“mark all read”, “delete all”)

A practical loop: queue 5–10 articles before your commute, blow through them on the train.

Voice recommendations

• Hard news → 阿井田 茂 (broadcaster baritone)
• Analysis pieces → 凛音エル (cool, controlled)
• Essays / blog posts → 桜音 or まい (warmer)
• Tech (Qiita / Zenn) → fumifumi / 阿井田 茂 (matter-of-fact)
• note long-form → みちのくあいり (expressive)

You can set a default voice per category in “Web Reader settings → Default voice”, so you don’t have to pick every time.

Speed sweet spots

• Skim listen → 1.5×
• Comprehend properly → 1.0×–1.2×
• Just the highlights → 2.0× + chapter skipping

Speed is remembered per article, so you can have “this one at 1.2×, that one at 1.8×”.

Audio export (v2)

“Want to share an article’s audio with a friend” / “want it as podcast material” — m4a / WAV export ships in v2. Commercial use is governed by per-voice ボイスモデル licence terms.

When things go wrong

• Body extraction fails on a site — 記事抽出 heuristics sometimes mis-classify. Send the URL and what you saw to [email protected] and we’ll investigate.
• Playback stalls — check device storage. 言語処理 model + a voice = roughly 1 GB needed.
• A character is consistently mispronounced — register it in the pronunciation dictionary. The override applies across every article.

Happy reading-by-ear.