Help & FAQs

Sign in with Google, finish onboarding (one minute), then on the home page pick a scenario - Casual, Interview, Workplace, Networking, Debate, Storytelling, Travel, or Reading aloud. The AI partner opens the conversation. Tap the mic and start talking.

Casual mode. It's the lowest-pressure way to warm up. The welcome screen you saw on your first visit takes you straight there. After 1–2 sessions, try Reading mode for pronunciation feedback - it's the most differentiated feature and the fastest signal of improvement.

Chrome, Edge, Firefox, and Safari - desktop and mobile. We use server-side speech recognition (Whisper via Groq), so you don't need a specific browser's STT support. iOS Safari and Chrome on Android both work.

No. It runs in the browser. You can "install" it on your phone home screen for one-tap access (Add to Home Screen on Safari/Chrome).

Yes — optional. The FluentWith Chrome extension translates any word or sentence on the web into your native language, looks up definitions, and explains tricky lines (Explain prefers Chrome's built-in on-device AI in Chrome 138+, and falls back to a server-side model when on-device isn't available and you're signed in). On YouTube, you can click any word inside a caption to look it up — the video pauses around the popover. Saving is optional; saved words flow into your spaced-repetition queue and AI speaking drills. It's free, works on Chrome / Edge / Brave / Arc, and you don't need it to use FluentWith — the main app works on its own.

When you tap the mic, your microphone records audio in your browser. When you tap the mic again (or auto-stop in Reading mode), the recording is sent to our transcription provider (Groq, using OpenAI Whisper) which returns text. The text is what we save and feed to the AI partner. Audio is transient on our side - we forward it to Groq and don't keep a copy.

Yes - but only on your device. When you read a passage aloud, the recording is saved to your browser's local storage (IndexedDB) so you can replay it on the report page ("Hear yourself" button). It never leaves your device except for the brief transcription round-trip. Sign-out wipes the local storage.

Whisper is accurate but not perfect - names, technical terms, and strong accents sometimes get misheard. Edit the text in the textarea before clicking Send. The AI sees your edited version, not the raw transcript.

So you can fix mistakes before sending. Edit freely, then hit Send (or Ctrl/⌘+Enter).

Browser mic indicators have a small grace period (1–2 seconds) after the stream is closed. If it stays on longer than 5 seconds, refresh the page - that's a bug, please email us.

5–15 minutes is the sweet spot. Sessions under 30 seconds skip the full report - there's not enough data for useful feedback.

Long sessions take 15–40 seconds to analyze. If it's stuck longer than a minute, hit Retry on the report page. The session is saved either way.

History page (top-right of the home page). Lists every ended session with scores, transcript, and links to the full report.

A short summary of your level + recurring patterns + goals that the AI partner reads silently before each turn. It steers the conversation toward your weak spots without breaking character. View it on the Progress page.

Read a passage aloud at the target speed. Whisper transcribes what you actually said; we compare it against the target text word-by-word. For mismatches (e.g., you said "tree" instead of "three"), we show the IPA (/θriː/) plus a plain-English mouth-shape instruction and an explanation of why STT misheard.

Yes - every reading session saves the audio locally on your device. The report page has a "Hear yourself" player and a "Correct version" reference voice so you can compare side by side.

On the reading session page, if transcription fails, a "Retry transcription" button appears. The recording is still on your device - we'll just re-send it. You don't need to read the passage again.

The teleprompter scrolls at the target words-per-minute for your level. Your actual WPM is measured from the transcript and recording duration - they should be close. If they're wildly off (e.g., teleprompter target 110 but you measured 145), the pacing maths might be off for that passage; report it via support.

After you make the same mistake category 3+ times across recent sessions. The system spots the pattern, generates a targeted drill, and queues it on the Drills page.

Dismiss it. We'll generate something different next time the pattern recurs.

Yes - Drills page → History (top-right). Shows every completed drill with your answer, correctness, and feedback.

Go to your account page and click Delete. Your data is marked for deletion immediately and hard-deleted within 30 days. You can restore within those 30 days by signing back in - we'll show you a restore screen.

Yes — email support@fluentwith.com from your account address and we'll send you a JSON file with every session, message, mistake, drill, attempt, goal, and snapshot we have for you. It does NOT include the audio recordings (those live on your device, not on our servers).

Sign out ends your session and clears your browser's local audio storage, but your account and history are kept on the server. Delete marks the account for permanent removal; after 30 days the data is wiped from our database.

No. See privacy policy.

No. We don't train models, and our AI providers (Anthropic via API, Groq via API) don't train on API content per their published policies.

India-based, governed by Indian law per our terms of service. EU/UK users still have GDPR rights; Indian users have rights under the DPDP Act 2023. Both detailed in the privacy policy.

Two: a strictly-necessary auth-session cookie (always on; cleared on sign-out), and Google Analytics cookies (off by default - only if you accept the cookie banner). Full list and opt-out: cookies page.

Click the "Cookie settings" link in the footer (or visit /cookies) and choose "Reject non-essential". Your choice persists across sessions on this device.