Privacy Policy
Last updated: 2026-10-08
In short: banter stores your account details, the transcript of what you said in a practice session (with word-by-word timing, so it can score your fluency), and your scores and usage. banter never stores a recording of your voice. Your microphone audio streams from your browser straight to Amazon Transcribe, which turns it into text, and after each thing you say, the audio of just that turn passes once through banter’s server to Microsoft Azure AI Speech, which scores your pronunciation and keeps nothing. The conversation text is processed by Amazon Bedrock (running Anthropic’s Claude models) to run the conversation and score it, and Stripe handles billing without banter ever seeing your card number. We do not sell your data or use it for advertising. You can export or delete everything banter holds about you from /account at any time.
Who we are
banter is operated by Idea Engine LLC, a Wisconsin limited liability company, referred to here as “banter”, “we” or “us”. We decide how and why your personal information is used (we are the “controller”). This policy covers anyone who visits banter, creates an account, or uses banter to practice a language.
Idea Engine LLC, Wisconsin, USAMailing address available on request at privacy@idea-engine.studioPrivacy questions: privacy@idea-engine.studio
What we collect
We collect only what the product needs to run a conversation, score it and bill for it:
- Account identity. You sign in with a one-time link we email you. We hold your email address. If we offer sign-in with Google and you use it, we also receive your display name, profile picture link and a stable identifier for your Google sign-in (through Amazon Cognito), and the sign-in tokens Google issues; we never see your Google password. There is no banter password at all.
- Account records. When your account was created and last used, whether you have used your one free conversation, and when you accepted these terms and which version you accepted.
- Sign-in records. While you are signed in, your browser holds a random session token in a secure, HTTP-only cookie, matched by a row on our side so we can sign you out everywhere at once. An emailed sign-in link is a single-use token that expires after 10 minutes and is deleted when used.
- What you say in a session. Each thing you say is stored as a text transcript, together with the timing of each word (when it started and ended, in milliseconds) and, where available, the speech recognizer’s confidence in that word. This is what the fluency metrics — speaking rate, pauses, response time — are calculated from. We also keep what the AI partner and the in-call coach said.
- Session choices. The scenario, language, level and conversation partner you chose, how long the session ran, and which version of our microphone notice you had acknowledged.
- Scores and corrections. The fluency score for each session, its CEFR band, the sub-scores that make it up, and any grammar, vocabulary, pronunciation or register corrections generated for that session, quoting the exact words you used. For each thing you said, the pronunciation scores Microsoft Azure returned (accuracy, fluency and completeness, a score for each word) and how many seconds of audio they measured — numbers, not the audio.
- Usage. How many minutes you have used today and this month, so your plan’s allowance can be enforced.
- Billing identifiers. If you subscribe, we store your Stripe customer and subscription IDs and your plan, subscription status and renewal date. We do not store your card number or bank details — Stripe holds those.
- Device and log data. Like any website, our hosting receives your IP address, browser type (user agent) and the pages and features you request with every request. We use this to deliver the service, keep it secure and prevent abuse. Our server logs, which can contain this information and error details, are kept for 30 days. We also count visits and sign-ins as anonymous totals (no identifiers) to see how the service is used. When a visit starts we also note the website that sent you (its domain only, for example google.com), any campaign tags in the link you followed (utm_source, utm_medium, utm_campaign, utm_content and source), whether that link carried a referral code (never the code itself), the page you arrived on, and whether the visit looks automated (a bot or crawler) or like a person's browser, which we work out from the browser type without storing it. This record does not include your IP address or any identifier, and we use it only to see which links bring people here.
What happens to your voice
Before your first live conversation, banter tells you where your voice goes and asks you to acknowledge it. When you press Start talking, your microphone audio streams directly from your browser to Amazon Transcribe over a short-lived, signed connection, and is turned into text in real time; the transcript and word timings are saved. Amazon processes the audio on our behalf under the AWS service terms. Under those terms AWS may store audio processed by Transcribe and use it to develop and improve its services unless the account has opted out; we are putting that opt-out in place for banter’s AWS account and will update this policy when it is.
To score your pronunciation, your browser also holds a rolling window of that same audio (never more than the last 40 seconds) in memory, and discards it when the call ends. When you finish saying something, the audio of just that turn (at most 30 seconds) is sent once to banter’s server together with the text Transcribe made of it, and passed straight on to Microsoft Azure AI Speech, which compares how you said the words with the words themselves and returns scores: accuracy, fluency, completeness, and a score for each word. banter saves only those numbers. The audio is held in memory for that one request and is never written to our database, storage, logs or metrics. Microsoft processes it on our behalf; Microsoft documents that it does not retain or store audio sent for pronunciation assessment, and banter does not switch on Azure’s optional audio logging.
banter does not identify you by your voice: it does not create voiceprints, does not use speaker identification or speaker recognition, and does not keep recordings. Pronunciation scores measure how closely your speech matches the words you said; banter does not use them to identify you. The optional video avatar does not change this — the avatar service receives only the words the AI partner is about to say, never your microphone — and banter never uses your camera.
Cookies and browser storage
banter uses only what it needs to work — no advertising or cross-site tracking cookies, and no third-party analytics:
- Cookies: your sign-in session cookie, and short-lived security cookies used during sign-in (a CSRF token and the page to return you to).
- Local storage (stays on your device): your light/dark theme, your chosen language, level and conversation partner, that you have read the microphone notice, and when you last visited (so a return visit is counted once).
- Session storage (cleared when you close the tab): that you checked your speakers and microphone this visit, and the short-lived credentials for a conversation you have just started.
Because these are all strictly necessary for the service, we do not show a cookie banner. You can clear them in your browser at any time; you will be signed out.
How we use it
- To run your practice conversations, score them and show your history and progress.
- To sign you in, keep your account secure and send you sign-in links.
- To enforce plan allowances and bill paid plans, and to email you about a failed payment.
- To keep the service reliable, investigate errors and prevent abuse.
- To comply with the law, including tax and accounting records.
We do not use your transcripts, recordings or scores to train AI models, and we do not use them for advertising.
Who we share it with
banter uses a small number of service providers (“processors”) to run the product. Each one receives only what it needs to do its job, on our behalf:
- Amazon Web Services (United States) — hosting and infrastructure: the website and servers (CloudFront, Lambda), the database (Aurora), logs and metrics (CloudWatch), and the services below.
- Amazon Transcribe (speech-to-text). Receives your microphone audio, and your IP address, directly from your browser during a live session.
- Microsoft Azure AI Speech (pronunciation scoring). Receives, from banter’s server, the audio of each thing you say in a live session (up to 30 seconds each) and the text of it, and returns pronunciation scores. It processes the audio in memory in the United States (Azure’s East US region) and, per Microsoft’s documentation for this service, does not retain or store it. It does not receive your IP address, your email or anything else about your account.
- Amazon Bedrock, running Anthropic’s Claude models (the conversation, the coach and the scoring). Receives the text transcript of your session — what you and the conversation partner said — to generate replies and, at the end, to judge your performance. Bedrock runs the models inside AWS: Anthropic does not receive your data, and Bedrock does not use it to train models.
- Anam (the video avatar, on plans that include it). Anam is a United Kingdom company. It receives the words the AI partner is about to say and your IP address (your browser connects to it directly to receive the video), and streams back the video of the partner saying them. It does not receive your microphone.
- Amazon SES (email). Delivers your sign-in links and billing notices. It receives your email address and the message itself.
- Stripe (payments). If you subscribe, Stripe receives your payment and billing details and processes the charge, as a processor for us and under its own privacy policy for its legal obligations. banter never sees or stores your card number.
- Amazon Cognito and Google (sign-in with Google, only if we offer it and you choose it). Google confirms who you are and releases your email, name and picture to Cognito with your consent.
banter’s code can also use HeyGen (an alternative video avatar), but it is not switched on. We will update this list before it receives any data.
We do not sell your personal information, and we do not share it for cross-context behavioral advertising. We may disclose information if the law requires it, to protect the rights, safety or property of our users or others, or as part of a sale or reorganization of our business, in which case this policy continues to apply.
Our legal bases for processing (UK and EU users)
- Contract. Most processing — running your sessions, scoring them, keeping your history, billing your subscription — is necessary to provide the service you signed up for.
- Legitimate interests. Security, abuse prevention (for example, detecting a compromised account burning through minutes), service logs and anonymous usage counts.
- Consent. Your microphone is used only after you acknowledge our notice and grant permission to your browser. You can withdraw it at any time by ending the session or revoking the permission in your browser settings.
- Legal obligation. Tax and accounting records.
Your rights
Wherever you live, you can:
- Access and export everything banter holds about you, including your full transcripts and scores, as a downloadable file.
- Correct your account details. Your email address is how you sign in, so to change it, write to us and we will move your account across.
- Delete your account and every row linked to it — sessions, transcripts, scores, corrections, usage records and your sign-in records. Deleting signs you out on every device immediately.
- Object to or restrict processing, and ask about the portability of your data to another service.
Export and deletion are self-serve, from /account. Deletion is immediate and irreversible once confirmed, and it also cancels any active subscription so you are not billed after your account is gone. For anything else, email privacy@idea-engine.studio from the address on your account; we will respond within 45 days. We will not treat you differently for using these rights. In the UK or EU you may also complain to your data protection authority.
US state privacy rights
Residents of California, Colorado, Connecticut, Virginia and other US states with consumer privacy laws have rights to know, access, correct and delete personal information, and to opt out of its sale, sharing for targeted advertising, or profiling. We offer access, correction and deletion to everyone, as described above. We do not sell personal information, share it for cross-context behavioral advertising, or use it for profiling that produces legal or similarly significant effects, so there is nothing to opt out of. We do not use your information to infer characteristics about you beyond scoring your language practice. To make a request, email privacy@idea-engine.studio; an authorized agent may make a request on your behalf with your written permission. We may need to confirm the request comes from the account’s email address. If we decline a request, you can appeal by replying to our decision, and if the appeal is denied you may contact your state attorney general.
How long we keep things
- Transcripts, scores, corrections and session records: for as long as your account exists, so your history and progress stay available to you. Deleting your account deletes them.
- Sign-in links: until used, or 10 minutes. Sign-in sessions: until they expire or you sign out or delete your account.
- Server logs: 30 days.
- Billing records Stripe holds on our behalf: as long as tax law requires, typically around seven years, even after you delete your banter account. That is a legal obligation, and it is why account deletion detaches your Stripe record from your banter account rather than erasing it.
Security
All traffic to banter is encrypted in transit (HTTPS with HSTS). The database is encrypted at rest, is not reachable from the internet, and is backed up with point-in-time recovery. Access to production systems is limited to the people who run banter. Vendor keys never reach your browser; the browser gets only short-lived, single-use credentials. banter does not store card numbers, government ID numbers or passwords. No system is perfectly secure, but if a breach affecting your personal information occurs, we will notify you and any regulator as the law requires — for Wisconsin residents, within the time Wis. Stat. § 134.98 sets where it applies.
Where your data is processed
Idea Engine LLC is based in Wisconsin, USA, and banter’s infrastructure runs in the United States (AWS region us-east-1). Anam, the video-avatar provider, is a United Kingdom company, so on plans that include the video avatar, the partner’s lines and your IP address are transferred to Anam and may be processed outside the US. If you are in the UK, EU or elsewhere outside the US, your data is transferred to the US to provide the service. Where the law requires it, we rely on the European Commission’s Standard Contractual Clauses (and, for the UK, the International Data Transfer Addendum) or another lawful transfer mechanism in our agreements with our processors.
Age requirement
banter is only for adults: you must be 18 or older to create an account, and you confirm this when you sign in. banter is not directed to children, and we do not knowingly collect personal information from anyone under 18 (or from children under 13 at all). If you believe someone under 18 has an account, contact us at privacy@idea-engine.studio and we will delete it.
Changes to this policy
If we change what we collect or who we share it with, we will update this page and the date at the top. For a material change, we will also email you at the address on your account before the change takes effect.
Contact
Questions about this policy, or a request that is not covered by the self-serve tools above: privacy@idea-engine.studio. General support: support@idea-engine.studio.
Idea Engine LLC, Wisconsin, USAMailing address available on request at privacy@idea-engine.studio