Privacy Policy

Babelarc is a real-time chat-translation overlay for gamers. This website (babelarc.app) and the desktop app are operated by the project maintainer. Reach us at babelarcapp@gmail.com.

• Account info — when you sign in with Google or an email link, we receive your email address, display name, and avatar URL (whatever the sign-in provider hands us). We never see your password.
• Usage records — for each translation we record which tier you used (Fast / Quality), the timestamp, how much energy was charged, and a request ID. Used for billing and troubleshooting.
• If you don't sign in, the app never talks to our servers at all, so we know nothing about you. Sign-in is required to use the translation feature.

• The chat screenshots themselves (we don't retain the originals on the Babelarc side long-term — but see §4 for edge-level caching and logging at our infrastructure provider)
• No local translation cache on your machine — the Babelarc folder doesn't write translation content to disk
• Game memory or process — we only capture the screen region you frame; we don't read game memory or inject/hook into the game
• Your keystrokes, or anything else on your screen
• Device fingerprints, browser history, or long-term IP logs
• Your social accounts or contact lists

When you press the hotkey, the app sends just the rectangle you framed over an encrypted connection to our infrastructure provider (Cloudflare), which forwards it to a third-party AI translation service. The translated text travels back the same way and lands in the overlay window.

We need to disclose the following honestly: to keep the service running, requests and responses can be briefly cached or logged at the edge of our infrastructure provider — including the framed image and the returned translation:

• Edge caching: if the same content is requested again, the cached response may be returned. The cache window is up to about 1 week, auto-purged after that.
• Operational logs: request and response payloads (including content derived from the framed image) may be retained in a capped log for service operation. Older entries are evicted when the cap is reached.

These two retention layers exist for infrastructure operation only — they are not used for model training and not used for advertising. On the Babelarc side we don't keep the image or translation content long-term; we only keep billing/diagnostic metadata (tier, timestamp, energy debit, request ID). On your machine no local translation cache is written — see the next section.

In addition: to improve translation accuracy, the third-party AI translation service may invoke its search capability while processing your request — to look up real-time information or verify terminology, e.g. new in-game terms, name spellings, or cultural references. Search queries can contain content fragments related to the translation (for example keywords or a chunk of context derived from the framed area). These searches are initiated autonomously by the AI translation service, and the search provider's retention policy is set by that provider. We do not separately send your data to any other third party.

Babelarc is a portable app — the entire Babelarc/ folder is the program, and all user data lives under Babelarc/data/. None of this ever leaves your machine. Deleting the folder is a complete uninstall. What's in there:

• auth.json — the sign-in credentials issued to you after login (so you stay signed in between launches) and your email. Stored in plaintext (a deliberate trade-off so you can copy the folder to another machine and stay signed in, instead of binding the credentials to your Windows account).
• settings.json — your preferences: hotkey, target language, appearance, saved framing regions, etc. Saved regions store rectangle coordinates and the names you give them, nothing about the actual screen content.
• logs/ — runtime logs for troubleshooting. No screenshots, no translation content — only operational events.

Each provider has its own privacy policy and a data-processing agreement with us.

• Sign-in — if you choose one-tap Google sign-in, Google handles that step; the email-link option doesn't involve Google at all. We only receive the email / display name / avatar the sign-in returns.
• Cloudflare — runs our website and translation API, stores your account and energy data, and distributes the app downloads, processed across its global network close to you.
• A third-party AI translation service — the translation model runs there; it processes the single frame / text you send and returns the translation.
• Creem.io (Glow Finance LLC) — handles all subscription billing as our Merchant of Record. We share only your email and an internal Babelarc user id with Creem so they can tie payments back to your account.Card numbers, bank details, and any payment credentials are collected and stored by Creem directly — Babelarc's servers never see them.

Because Creem is the Merchant of Record, your card statement shows "CREEM.IO" rather than "Babelarc"; Creem also handles global tax compliance and chargebacks. See Creem's own privacy policy at creem.io/legal/privacy.

The website uses a single functional cookie to remember your language preference — no ads, no user tracking, no third-party analytics.

Once you subscribe, our server keeps the following on your account record:

• Current tier (Starter / Pro / Power)
• Subscription status (active / canceled / expired / etc.)
• Current period start and end timestamps
• Creem subscription id and customer id (used to look up your subscription for the manage / cancel flows — not a payment credential)
• A scheduled plan change, if any (e.g. if you've scheduled a switch from Power to Pro at the next renewal, we record the target tier so the change applies automatically at period end)
• Energy grant / debit ledger entries (auditable, immutable)
• A short-lived concurrency-guard flag (used to prevent accidental double-charges when you click Subscribe on both the client and the website at the same time; expires automatically within minutes)

These exist so we can reconcile and so you can inspect your subscription state. The actual card information, cardholder name, and billing address stay with Creem — Babelarc never sees or stores them.

Server-side translation usage records are kept for 90 days, then aggregated and anonymised — we keep the totals ("how many translations ran"), but lose the link to any individual user. Account info is kept until you ask us to delete it; once you ask, we erase every field that could identify you within 30 days, leaving only the aggregate stats. Local data (the files listed in section 5) is yours to control — deleting the Babelarc/ folder wipes everything.

You can ask to see what we have on you, correct it, export it, or delete it. Email babelarcapp@gmail.com from the address you signed in with and tell us what you want. We respond within 30 days.

When we make material changes to this policy, we'll notify you inside the app and by email before they take effect. Continued use after that point means you've accepted the new version.