Record your private clone in any supported reference language, then use that same voice for English, Chinese, German, Spanish, French, Italian, Japanese, Korean, Portuguese, or Russian TTS.
Yes—one CastReader voice clone can generate TTS in all 10 currently supported output languages: English, Chinese, German, Spanish, French, Italian, Japanese, Korean, Portuguese, and Russian. The language you choose while recording only changes the reference passage; it does not lock the clone to that language. After creation, choose the output language separately, paste up to 600 characters for a short WAV, or select the clone in the document reader for longer authorized text. Pronunciation, accent, and cross-language naturalness can vary, so review names and learning material before relying on the result.
The clone preserves voice identity across supported languages; it does not replace a native teacher or pronunciation review.
Choose an output language, paste a reviewed sentence, generate it, and download the short WAV.
Generate a language clipOpen an authorized document, choose your owned or shared clone, and keep the text visible while listening.
Open study materialAsk the living voice owner to create a private, expiring Voice Gift for your exact account.
See Voice GiftTry the voice library when a native-style built-in voice is more useful than familiarity for the exercise.
Browse voicesThree independent choices make the multilingual workflow easier to understand and review.
Choose a passage you can read naturally for the 3–30 second recording. It does not restrict later output languages.
Select one of the 10 supported output languages for the current text-to-speech request, independent of the recording prompt.
Check spelling, names, stress, meaning, accent, and pacing. Generated speech is practice support, not certified pronunciation instruction.
Keeping these layers separate avoids the common assumption that a clone only speaks the language used during recording.
| Layer | What it controls | What it does not guarantee |
|---|---|---|
| Recording passage language | The text you read while creating the voice | A permanent language limit |
| Output language | The synthesis mode for the current request | Native accent or perfect pronunciation |
| Text you provide | The exact learning or listening content | Accuracy, translation quality, or age suitability |
| Voice identity | The familiar timbre carried into output | That the real person speaks or endorses the generated words |
Choose one of the 10 reference passages, record your own voice for 3–30 seconds, and confirm ownership and consent.
After the clone is ready, select the language for the text you want to hear. It can differ from the recording passage.
Start with up to 600 characters and listen carefully to names, unfamiliar sounds, numbers, and sentence endings.
Download the approved WAV, or open longer authorized material and select the same clone in the web document reader.
A familiar voice can lower the emotional distance of a new language. Learners may be more willing to replay a sentence when it sounds like themselves, a parent, or another living person who deliberately shared a Voice Gift. The value is continuity of voice, not a claim that the clone becomes a native speaker.
Cross-language voice cloning separates who the voice resembles from which language the current request speaks. CastReader offers 10 reference passages and the same 10 output-language choices; choosing Chinese for the recording prompt does not prevent a later Spanish or English request.
Generated pronunciation can vary by voice sample, language pair, text, names, and punctuation. Use short tests, compare important material with a trusted dictionary or teacher, and do not treat the audio as authoritative instruction for high-stakes communication.
The generated words are not a statement made or endorsed by the real speaker. Use your own voice or a valid Voice Gift, keep the purpose personal, and disclose AI-generated speech when a listener could otherwise misunderstand its origin.
What one clone can do across languages—and what still needs human review.
No. One private clone can be selected for all 10 supported output languages. The recording-language tabs only provide reference passages you can read naturally.
The current cloned-voice output choices are English, Chinese, German, Spanish, French, Italian, Japanese, Korean, Portuguese, and Russian.
Not necessarily. Cross-language pronunciation and naturalness vary. Review every important result, especially proper names, abbreviations, numbers, tonal distinctions, and unfamiliar words. A built-in voice from the voice library may be a better comparison for some exercises.
Yes. The short voice-cloning generator accepts up to 600 characters and provides WAV download after generation. Longer authorized text can use the document reader and its current bounded workflow.
Only with the living voice owner’s active permission. Ask them to create their own clone and a recipient-bound, expiring Voice Gift; do not upload their recordings yourself.
Create your private clone and generate a short clip in a supported output language.
Open longer authorized material and select an owned or shared cloned voice.
Receive a familiar voice from a living owner through a named, expiring grant.
Compare the cloned result with CastReader’s available reading voices.
Download the CastReader app for DRM-free EPUB import and supported mobile reading workflows.
Hard numbers — not vibes — from authoritative sources
$2.22 billion
US audiobook sales in 2024, up 13% year-over-year (Publishers Weekly / Audio Publishers Association)
Source →51%
of US adults have listened to an audiobook in 2025 — roughly 134 million people (APA Consumer Survey 2025)
Source →2.2 billion
people globally with near- or far-vision impairment (WHO Fact Sheet, 2024). TTS is the primary access path for digital reading content.
Source →78%
of audiobook listeners multitask while listening — commute, chores, exercise (Audiolibrix Great Audiobook Survey, 2024)
Source →27.2 minutes
average single-trip US commute in 2024, up from 26.8 (US Census ACS via Statista). That's nearly an hour each day of audio-only time.
Source →effect size 0.35
measured comprehension lift from TTS for reading-disabled students across 22 studies (Wood, Moxley, Tighe & Wagner, Journal of Learning Disabilities, 2018)
Source →15.5 million
US adults with ADHD per CDC 2024 — about half diagnosed in adulthood (CDC MMWR, October 2024)
Source →Every Chrome Web Store review below is verifiable at the link in each card. We don't hide negative feedback — we answer it within 24 hours.
“This is 1 of the best TTS and its smooth. If this is truly free i'll keep this 100%. Every other TTS says its free but has a secret. They interrupt or they just say better ai voices pay. But i like this voice. I've tried loads and this is 1 of the best ones that actually says free.”
“Works perfectly on vivaldi. One suggestion though. I wish it had a play button appear next to a paragraph when we hover over it. Just like in the case of speechify.”
“Extremely user friendly short keys. Placed forward backward and speed up down as Natural as it could be. Voices are great and smooth. I would recommend it over many hyped products.”
“At the very least it's better than many paid TTS models. Still not as good as ElevenReader or LAP, but maybe the best free model for TTS.”
“So glad I can finally switch voices! The default was fine but I found one I actually enjoy listening to for hours. Small thing, huge difference.”
“Best one i found, user friendly, and great voice over.”
“ChatGPT's long answers are finally listenable. Let it generate while I listen — doubles my productivity. Love the inline button next to each response.”
“I tried using this add-on to listen to an ebook on the O'Reilly learning platform, and it works smoothly. However, it always restarts from the first paragraph whenever I scroll or select a different paragraph. Please consider adding a bookmark or checkpoint feature so users can mark where the reading should begin.”
↪ Founder reply
Replied by CastReader founder Yan Xu within 48 hours: acknowledged the issue, shipped a bookmark feature in the following release. Reviewer's verbatim feedback drove the v1.2 roadmap.
“Need to highlight text and select it.”
“Hard to select text.”
↪ Founder reply
Replied by CastReader founder Yan Xu within 24 hours: apologized, asked which site/browser the issue occurred on, provided a workaround using the keyboard shortcut, and offered direct support at support@castreader.com.
We re-test, re-write, and ship continuously. Every entry has a real date.
Rewrote landing pages with verbatim Chrome Web Store testimonials, real audiobook market data, and tested-12-extensions methodology. Every claim now has a sourceable link.
CastReader provides native apps for iPhone and Android; Android downloads are available directly from Google Play.
Wrote up the OCR pipeline: how CastReader handles Amazon's 184 random font alphabets and 361 unique glyphs per Kindle book. Shared in dev.to.
The phone app flow now has its own landing page for bookshelf sync, Read Aloud, synced highlighting, auto-scroll, background playback, and resume progress.
Ranked #10 in Daily, 99 upvotes, 4 community comments shaped the v1.2 roadmap.
Switched from older TTS engines to Kokoro neural voices. User reviews shifted from 'usable but robotic' to 'enjoy listening for hours' (verbatim from review by patrick chiang).
OCR success rate improved from 78% to 89% on English-language Kindle books. Multi-column page detection added for academic PDFs.
I built CastReader because I owned hundreds of Kindle books and couldn't listen to them on my morning runs without buying separate Audible copies. The technical problem — Amazon's Cloud Reader font encryption — turned out to be solvable with OCR. The product problem — making it actually pleasant across phones, desktops, and 9 supported languages — took two years of iteration. We're a small team. I answer every Chrome Web Store review personally (see testimonials above — including the 3-star and 1-star ones). If something's broken or missing, email support@castreader.com.
— Yan Xu, founder
Last reviewed: · CastReader Team — reviewed against 2025 testing data
Record once, choose a supported output language, and generate a short practice clip in the same private clone.