Frequently asked questions (FAQ)
Overview
Here are short answers to the most common questions about Whisperer — from privacy on a call to billing and choosing a model. The answers are brief (1–3 sentences) and link to detailed articles if you need to dig deeper.
The questions are grouped by topic: general product info, installation and permissions, audio and video services, plans and billing, AI and models, the knowledge base, privacy and no-logs, and account.
When to use
- You have a short, specific question and need a quick answer.
- You're just getting to know the product and want to clear up typical doubts.
- You're looking for which article to check for details on a topic.
Step-by-step
To find an answer: 1) identify the topic of your question; 2) go to the corresponding section below; 3) if needed, open the related article via the link in the answer.
General / Product
What is Whisperer? A personal AI call assistant: an overlay on top of any video service that listens to the audio, keeps a transcript, and suggests answers via an LLM. More — What is Whisperer.
Which video services does it work with? Any of them — Zoom, Google Meet, Microsoft Teams, Telegram, Discord, and others. There's no integration into the service: Whisperer works as an overlay on top of the app.
Do I need to invite a "bot" into the call? No. Whisperer doesn't join the call as a participant and doesn't require an invite link — it runs locally on your Mac.
Can the other person see the suggestions and transcript? No. The overlay is invisible in broadcasts and screen sharing — only you see it.
Who is Whisperer for? Job seekers (including System Design interviews), teachers and tutors, sales and managers, and anyone who's often on calls.
What is System Design mode? A mode in which the model produces a structured answer with diagrams (Mermaid). Available on paid plans.
Does Whisperer translate the conversation? Yes, translation can be enabled in the overlay settings; it's shown as a separate line (TranslationStrip).
Installation and permissions
Where do I download the client? Distributions are downloaded from the web dashboard (the download section); the list of builds with version and size is served by the builds service.
Which OS does Whisperer run on? There are native clients for macOS (Intel and Apple Silicon) and Windows (10 version 1809+ or 11). The web dashboard (history, knowledge base, billing) is accessible from any browser. See Installation on macOS and Installation on Windows.
Is there a Windows version? Yes — it's a full-featured native app with capabilities close to macOS (live transcription, overlay, screenshots, translation, hotkeys). It needs Windows 10 1809+ or 11 and the WebView2 Runtime (installed automatically). See Installing the client on Windows and Permissions on Windows.
Which macOS permissions are needed? Microphone (your voice), Screen Recording (the other person's system audio and screenshots), and Accessibility (global hotkeys).
What about on Windows? Only the microphone. The other person's system audio is captured automatically; screenshots and hotkeys work without separate permissions. See Permissions on Windows.
Why does it need the "Screen Recording" permission if I'm not recording video? macOS uses it to provide system audio (the other person's voice) and screenshots. Without it, the assistant won't hear the other person.
What should I do if permissions aren't granted? On first launch the permissions wizard (PermissionWizard) starts and guides you to System Settings → Privacy & Security.
How do I set up hotkeys? Five actions (pushToTalk, toggleOverlay, searchKB, captureScreenshot, textInput) are configured in the client and in the web dashboard.
Audio and video services
How does Whisperer tell who's speaking? It captures two sources: the microphone ([Me]) and system audio ([Other]), and tags each fragment with a speaker.
The assistant can't hear the other person — why? Most likely the "Screen Recording" permission isn't granted: it's exactly what provides access to system audio. Grant it in System Settings.
Can I upload a ready-made audio file for recognition? No. Transcription is streaming only (real-time); there's no batch file upload.
What language is the transcript in? The language is set per session (Russian by default). Recognition is multilingual (Whisper, standard language codes).
How do I change the transcription language? The language is chosen at session start (in the hello parameters); the default is ru, and you can specify another supported language.
Is there noise suppression? Yes, noise suppression is enabled in the overlay settings along with width, opacity, and font size.
Can I work from screenshots? Yes. There's single and multi-screenshot (accumulating frames) for analyzing images in vision mode; multi-screenshot is on paid plans.
Plans and billing
How does Free differ from the paid plans? Free gives 60 minutes per month and basic suggestions, but without the core — no sessions with history, knowledge base, multi-screenshot, System Design, or frontier models. More — Subscriptions, plans, and quotas.
What plans are there? Free, Start, Pro, and Max. Differences in capabilities are in the article on subscriptions; prices and periods are on the plans page.
What does Max include beyond Pro? Access to frontier models (flagship Claude Opus, top-tier GPT, and equivalents). The rest of the core is the same as in Pro.
What are frontier models? The most powerful LLMs; in Whisperer they're unlocked only on Max plans.
How many minutes on Free? 60 minutes per month. On paid plans minutes are unlimited.
What happens when the minutes run out? The limit is soft: the current session will finish, but you can't start a new one until the quota is restored or you switch to a paid plan.
What currency do I pay in? The currency is determined by country: Russia — rubles, Europe — EUR, the rest of the world — USD. Details on the plans page.
How do I pay in EUR or USD? EUR/USD payments go through moneta.ru (PayAnyWay; Visa/Mastercard/Maestro); ruble payments — through TBank (MIR cards).
Why is the currency price higher than a direct conversion from rubles? A 1.30 multiplier is applied to the base ruble price and divided by the Central Bank rate, rounded down; the markup is not applied to discounts and bonuses.
How do I get a refund? Refunds are processed through support — there's no self-service cancellation in the dashboard. Write to support.
What is a promo code? A discount code: type percent (a discount percentage) or bonus_days (bonus subscription days).
How does the referral program work? For each qualifying purchase by an invitee (Pro or Max) you get 500 ₽; one reward per invitee.
AI and models
Which model recognizes speech? Whisper, real-time only.
Which LLMs are available for suggestions? Model families: Claude (Opus/Sonnet/Haiku), GPT, DeepSeek, Qwen, Gemini, Grok, Llama. The set depends on your plan.
What is a "model role"? The model's purpose: responses (answers), coding (code), vision (screenshots), system_design (diagrams), generation (mind-map), transcription (speech), embedding (RAG). You can set a separate model for each role.
How do I switch the model? In the settings you can assign a model to each role; the available set is limited by your plan (frontier — only on Max).
What is the model's answer made up of? The system prompt: role prompt → user context → the knowledge-base block (RAG).
What is a meeting mind-map? A meeting map of topic / action_item / decision nodes that Whisperer extracts after a session; it can be further enriched in the dashboard.
Knowledge base
What is the knowledge base? A store of your notes (Markdown + tags) whose facts are mixed into the model's answers via RAG. Available on paid plans.
How do notes get into answers? A note is split into chunks, turned into embeddings, and stored in a vector database; for each query cosine search finds relevant fragments and inserts them into the prompt.
What do wiki-links [[...]] do? They connect notes into a graph (neighbors, shortest path, centrality) — handy for navigation and search.
Who can see my notes? Only you. Access is closed on a fail-closed basis.
How do I refresh search after edits? Run reindexing (via the button/reindex) — it rebuilds the embeddings in the background.
Privacy / No-logs
What is no-logs mode? A session mode in which the transcript and answers aren't saved and are deleted after the session ends. More — Security and privacy.
Does no-logs spend minutes? Yes. Minutes are spent as usual; only data storage is saved.
What's stored after a regular session? The transcript and answers (for history and search), except for no-logs sessions. The knowledge base — until you delete it.
Where are the files stored? Locally on your device; there's no external object storage (S3).
How is access to my data protected? Data is isolated per account and tokens are stored in the system's secure storage. More — Security and privacy.
How do I delete my data? Sessions and notes are deleted in the dashboard; full account deletion — through support.
Account
What's needed to register? Name, e-mail, password (at least 8 characters), and passing the captcha. More — Registration and e-mail confirmation.
The confirmation code didn't arrive — what do I do? Check "Spam," wait ~60 seconds, and click "Resend." The code lives for 5 minutes and allows up to 3 attempts.
The code expired or I ran out of attempts. Request a new code (cooldown ~60 s); the new one again gets 5 minutes and 3 attempts.
Can I sign in before confirming my e-mail? No, sign-in is blocked until the address is confirmed.
How do I reset my password? Request a link on the sign-in page; it's valid for 24 hours and works once. Changing the password logs out all devices.
How do I pair the macOS client with my account (device code)? In the web, open the connection section — the client receives a one-time code (lives 30 minutes), which is exchanged for tokens; the redirect into the app goes via the whisperer:// scheme.
How do I change the interface language? Whisperer supports 16 languages; the choice is stored in your profile. On first login, the language is detected from your browser and geo.
What account statuses are there? active (active), frozen (frozen), banned (banned).
Screenshots
📸 [Screenshot: Whisperer overlay on top of a video-call window]
📸 [Screenshot: e-mail confirmation code entry screen]
📸 [Screenshot: plan and currency selection in the dashboard]
Common mistakes
- Waiting for a bot in the call. Whisperer works locally as an overlay; it doesn't join the call.
- Not granting "Screen Recording." Without it, the other person isn't heard and screenshots don't work.
- Considering no-logs free in terms of minutes. Minutes are always spent.
- Looking for an audio file upload. Transcription is streaming only.
Best practices
- Before your first call, go through the entire permissions wizard.
- Fill in your knowledge base and user context in advance — the answers will be more accurate.
- For sensitive conversations, enable no-logs.
- If you keep hitting the 60 minutes of Free — switch to a paid plan for unlimited minutes and the core.