Privacy policy
Wakaru Studios · last updated 7 September 2026
This page lists everything Wakaru sends off your machine, where it goes, and why. Subtitle text and the words you click leave your machine to make the features work. When you are signed in, alignment diagnostics include the episode ID, model, token usage, estimated cost, and processing measurements so we can check performance and costs. There are no ads, and nothing is sold or shared for marketing. We measure use of our own popup, settings page, and website separately; those PostHog measurements never include the Crunchyroll player. Details under Alignment diagnostics and Usage measurement.
Who we are
Wakaru is made by Wakaru Studios. For anything in this policy, email [email protected].
Your account
Signing in with Google gives us three things: your email address, a Google account identifier, and the time you signed in. That's the whole account record. We use it to run your free trial, check your subscription, and attach your synced data to you. We never see your Google password, contacts, or anything else in your Google account.
What the extension sends to our server
Our server (api.wakarustudios.com, hosted on Cloudflare) receives:
- Subtitle text — the Japanese and English lines of the episode you're watching, sent for AI word alignment and for the per-line explanation feature. Our server forwards them to OpenAI to compute the alignment, and the resulting word links are cached in a shared cache (keyed by the subtitle content, not by you) so each episode is only aligned once for all users.
- Alignment diagnostics — when signed in, the extension sends account-associated episode IDs, model configuration, token usage (including cached input and reasoning when reported), estimated cost, duration, time to first word links, retries, completion status, and extension version. These measurement reports do not include subtitle text or word-link output.
- Single words — when you play pronunciation, the word is sent to our server, which forwards it to Cartesia to generate the audio. Generated clips are cached on your machine.
- Your word list and settings — if you're signed in, saved words (with their context sentence, show name, and scene picture) and your settings sync through our server so your devices stay in step.
- Feedback — if you use the feedback button, we receive your message and, only if you leave the visible checkbox ticked, diagnostics: extension version, browser version, the show and episode open, and the subtitle/sync/AI status. Anonymous feedback also records your IP address, used only to rate-limit spam and never shown alongside your report.
- The uninstall survey — the optional page that opens when you remove Wakaru sends only what you type, plus the extension version.
Services the extension talks to directly
Some lookups go straight from your browser to public services:
| Service | What is sent | Why |
|---|---|---|
| jimaku.cc | Show and episode names | Finding Japanese subtitle files |
| AniList | Show names | Matching a title to its alternate names |
| jisho.org | A word you clicked | Dictionary lookup for words our bundled dictionary lacks |
| JapanesePod101 | A word and its reading | Native-speaker pronunciation recordings |
Each of these operates under its own privacy policy. None of them receives your name, email, or account details from Wakaru.
Alignment diagnostics
Cloudflare stores alignment reports for 90 days. Reports are associated with the signed-in account and identify the episode whose subtitles were aligned. They contain model, token usage, cost, timing, completion status, and counts. Reports distinguish fresh model calls from cached results. Costs are estimates or provider-reported charges, not billing invoices. These measurements are not accuracy scores.
These reports do not contain provider API keys, session tokens, browser history outside the alignment episode, video, audio, or screenshots. They go to Wakaru's server, not PostHog or advertisers. Measurements are held in memory and sent when alignment ends; no diagnostic history or upload queue is saved in Chrome. An offline connection or browser shutdown can prevent a report from arriving. Normal subtitle and alignment caches remain on the device to support playback.
Usage measurement
To see whether Wakaru is easy to use, especially for someone who just installed it, we record how our own screens are used: the extension popup, the settings page, and the pages of this website. This is done with PostHog, a product-analytics service, with data stored on its US servers. What it receives:
- Clicks and page views on those screens, with the label of what was clicked and when.
- Session replays — a reconstruction of those screens as you used them, so we can watch where people hesitate. Anything you type is blanked out before it leaves your browser; API keys and passwords are never recorded.
- Basic technical details — browser and extension version, screen size, country.
- A random identifier stored in your browser so repeat visits count as one person. It is not linked to your email.
None of this runs on Crunchyroll. The subtitle overlay sends nothing to PostHog, so which shows you watch, which words you look up, and what you save are never part of it. Replays are kept for 30 days. If you would rather not be counted, a content or tracker blocker that blocks posthog.com stops it entirely, and nothing in Wakaru breaks.
What we never collect
- No PostHog tracking on Crunchyroll — that usage measurement covers Wakaru’s own screens only. The separate alignment diagnostics described above identify alignment episodes.
- No general browsing history — we do not collect pages you visit outside the episodes involved in alignment reports.
- No video, audio, or screenshots of what you watch (scene pictures on saved words come from Crunchyroll's public preview thumbnails and stay in your own synced storage).
- No payment card details — payments will be handled by a payment processor, and card numbers never touch our servers.
This website
The website sets no cookies. It runs the same PostHog usage measurement described above (page views, clicks, session replay with typed text blanked), stored in your browser’s local storage rather than a cookie. If you sign in on the account page, a session token is kept in your browser's local storage so you stay signed in; signing out removes it. Headline fonts load from Google Fonts, which means your browser requests the font file from Google.
Where data lives and how long
Accounts, synced data, feedback, and alignment diagnostics are stored with Cloudflare (our hosting provider). Account-associated alignment reports are retained for 90 days; new reports do not store subtitle-output snapshots. Snapshots collected before September 23, 2026 expire under their original 90-day retention period. Usage measurement data is stored with PostHog in the United States. Shared alignment caches are separate: they are keyed by subtitle content and kept indefinitely, and contain subtitle text and word links without account details. Your account and synced data are kept while your account exists.
Deleting your data
Email [email protected] from your account's address and we'll delete your account, word list, settings, images, and account-associated alignment reports. Historical subtitle-output snapshots contain no account details and expire as described above. Words saved only on your device can be deleted right in the extension.
Changes
If this policy changes in a way that matters, the date above changes with it, and significant changes are announced in the extension before they take effect.