Answers to your questions.

Kalima turns live speech or recordings into accurate, searchable text. You get speaker labels, live translation, and AI summaries. No bot ever joins your call.

There's a free plan that opens without a credit card, so you can run a real meeting first. Sign up, open Kalima in your browser or the desktop app, and start a session that captures your mic, your desktop audio, or a meeting. Words appear as they're spoken, and paid plans add more minutes, sharing, and collaboration.

You can sign up with your email and a password, or use your Google account. For an email password, use at least 8 characters with an uppercase letter, a lowercase letter, and a number.

You don't have to install anything. Kalima runs right in your browser, and there's a desktop app if you prefer one. Either way it listens from your own machine, so nothing joins your call as a guest.

Yes. There's a desktop app with installers for Windows, macOS, and Linux. You can also add Kalima to your home screen or dock as an app and run it like a native one.

Kalima detects the spoken language on its own, with no setup, and live translation works across 60+ languages. It even keeps up when someone switches language mid-sentence, and several languages can appear in a single session.

No. Kalima listens right from your machine, in the browser or the desktop app. Nothing shows up as a guest in your call. There's no awkward join, and nothing for IT to approve.

Any of them. Kalima records the audio on your own device instead of plugging into each app. That covers Zoom, Google Meet, Microsoft Teams, and any other meeting tool, plus in-person conversations in the same room.

Yes. Kalima can record your computer's audio and your microphone together, mixed into one stream. That way you catch your own voice and everyone else's in the same transcript.

Yes. Kalima detects each speaker and labels them automatically. The labels stay consistent as Speaker 1, 2, and 3 across the whole session, even if you stop and start several times. You can rename anyone, and those names carry through the transcript and your exports.

Yes. The current sentence appears in real time with a LIVE indicator before it's finalized. Finished words stream in as people speak, and you can click any word to jump straight to that moment in the audio.

You won't lose the words. Audio keeps buffering on your device through the outage, then the part that dropped is re-transcribed and merged back into the transcript. Lines that were recovered after a gap are marked, so you can see exactly what was reconstructed.

On Windows it can. Once you enable it, the desktop app watches for active meeting apps and starts recording on its own. You stay in control: turn detection on or off, show an assist overlay with record and dismiss buttons, allow background recording, auto-stop when the meeting ends, and switch detection off for specific apps.

Yes. Kalima translates live, as the conversation happens. The translation streams in alongside the transcription and appears right next to the original text, so you read both in real time.

Live translation supports 60+ languages. There's no cap on how many run at once in a single session: languages per session are unlimited on every plan.

One-way translates everything into a single target language. Two-way handles a bilingual conversation, transcribing and translating both directions at once, with the original and the translations shown together in the live transcript.

Kalima handles it. Each word carries its own source language, and the transcript splits on speaker and language changes. A switch mid-sentence is detected and shown correctly instead of getting mangled.

Yes. In the studio you can place the original on the left and the translation on the right, or stack them, with a toggle. You can also filter the view to show everything, the original only, or the translation only.

Yes. The translation travels with the export: JSON includes the translation for each segment, and plain text shows it as indented arrow lines. For subtitles, you pick the original or translation track (the dropdown appears once translations exist). You can also share a translated session by link, grant, or invite.

Yes. Drop in an audio or video file you recorded earlier and Kalima transcribes it. Drag and drop works, and you get back a timestamped transcript.

You can upload MP3, WAV, M4A, MP4, MOV, AAC, FLAC, OGG, Opus, WebM, AIFF, and AMR. Video files are fine too: only the audio track is transcribed.

It depends on your plan. The largest single file ranges from 100 MB on the entry plan up to 1 GB on higher plans, and each file can run from 60 up to 300 minutes by plan. There is also a hard ceiling of 300 minutes (5 hours) per upload, checked before the file is sent. Anything longer is rejected with a clear message.

Your video container never leaves your device. Kalima pulls the audio out of the video in your browser first and sends only that. Before any upload, the audio is also compressed to a small, seekable file (about 28 MB per hour), so even long recordings move quickly.

Once the file has finished uploading and processing has started, yes. The job runs on Kalima's servers and is not cancelled, and the session stays in a processing state so you can come back later and check it. If you close the tab while the file is still uploading, that in-flight upload is cancelled, so you would need to start it again.

It depends on the length of your file, so there is no fixed number: longer recordings take longer. While it runs you see live progress: percentage, time remaining, and a stage tracker. The finished transcript comes back timestamped and split by speaker and language, with speaker labels and automatic language detection. You can also translate it during upload into 60+ languages.

Once a session is transcribed, Kalima can summarize it, answer questions about it, clean up speech-to-text mistakes, name the session for you, and read any line aloud. Summaries, chat, autocorrect, auto-titles, and read-aloud all work from the same transcript.

Yes. Kalima writes a structured summary with an Overview (a short TLDR plus key takeaways), a Tasks view of action items, and a Topics view, with a plain markdown copy available too. It adapts to the session: for a multi-speaker meeting it pulls out decisions, action items, and open questions, while for a solo talk, lecture, or podcast it captures the ideas and takeaways instead of forcing tasks. Each action item can carry an owner, a due date as it was spoken, and a high, medium, or low priority.

Yes. There is a chat that answers from the session content, and it tells you clearly when an answer is not in the transcript. Your conversation is saved per session and reloads when you reopen it, so your back-and-forth is kept. Chat history is private to you: someone viewing the session through a share link sees the transcript but not your chat. The chat stays on topic and declines unrelated requests like writing code, homework, news, or the weather.

Yes. With AutoCorrect+, you right-click a word or phrase in the transcript and mark it with a category like wrong word, name or term, grammar, punctuation, or filler. Each mark then becomes a fix you can resolve: type the replacement yourself, or use Fix with AI to suggest a correction for that span, with an undo stack so every change is reversible. When you correct a name or term, Kalima can find its other occurrences across the transcript and check each one in context before replacing, so it only changes the ones that are genuinely wrong.

Yes. When a recording ends, Kalima writes a short descriptive title (up to eight words) from the transcript and names the session for you, in place of a generic timestamp. It happens automatically and it is free on every plan, so there is nothing to trigger and no quota to spend. For read-aloud, right-click a segment to Speak Original or Speak Translation, and you can adjust the playback speed. It uses your browser's built-in speech for instant playback and falls back to a server voice for languages or browsers it can't handle on its own.

The metered AI features each have their own quota: separate allowances for summaries, chat messages, and autocorrections, set by your plan. If you use one up, that request pauses with a Quota Exceeded message until it resets the following week or you upgrade. Automatic session titles are free on every plan and do not count against any quota. There is also a shared rate limit so requests are not hammered, and a rate-limited request tells you when to try again. On quality, the summary is built to never invent decisions, owners, dates, numbers, names, or quotes that the transcript does not support, and to leave a section empty rather than pad it with filler. Your transcript is also treated as untrusted data, so the AI will not follow instructions hidden inside it.

Yes, in the Studio editor. You select a word or phrase, mark it for correction, then resolve the mark in the AutoCorrect+ panel: you can type the replacement text yourself, let AI suggest a fix, or apply the same fix everywhere it appears. You can undo your last mark with Ctrl+Z or Cmd+Z, undo an applied correction, and clear the whole transcript with a confirmation step.

Yes. Speakers are detected automatically and numbered across the whole session (Speaker 1, 2, 3, and so on), each with its own color. As the session owner, you click a speaker badge to give them a real name. The name saves and shows in the badges, segment headers, and exports, and any speaker you don't name falls back to 'Speaker N'. Only the session owner can rename, and during a live session the new names are pushed to everyone watching in real time.

You can export to PDF (a printable styled document that opens in a new window), SRT subtitles, VTT subtitles, JSON, and a Markdown summary, with a plain-text export also available. SRT uses standard SubRip timecodes and VTT uses WebVTT timing that works with HTML5 video players, and both can include speaker prefixes. The JSON export is the complete one: it carries each segment's text, translation, speaker, language, per-word timing and confidence, time ranges, and overall word and character counts. Your custom speaker names are included in every format.

Yes, for SRT and VTT. When a session has translations, the export gives you a track chooser so you can save the original transcribed track or the translation track. If there are no translations, the chooser simply doesn't appear.

Yes. You can style the on-screen captions with one of 7 built-in presets: classic, translucent, cinema, minimal, outline, high-contrast, and frosted glass. Your choice is saved and stays put across reloads. This styling applies to the captions as you watch them in Studio. Exported SRT and VTT files carry the text, timecodes, and speaker prefixes rather than the visual style.

Yes. Search runs over your session names, all transcript text, and all translations, then ranks results by relevance and shows highlighted snippets with the exact matching segments. It handles both whole-word and partial matches, so a partial term can still find the right session. It only looks through your own saved sessions, not deleted or unsaved ones.

Yes. Your audio and transcripts are encrypted in transit and at rest, and everything is EU-hosted. We never use your conversations to train AI models, and you can delete anything anytime. Kalima is GDPR compliant.

No. We never train AI models on your conversations. Your transcripts are yours, and we don't feed them into model training.

Yes to both. You can delete your data anytime, and you can also request a full export of what you have stored (GDPR data portability). If you delete your whole account, your transcripts and audio are permanently removed and your personal details are anonymized. Any active paid subscription is canceled, and you're signed out. Deleted sessions aren't kept forever: a scheduled background job purges them and clears the stored audio.

You stay in control. You can put a password on any share link, set it to expire on a date you choose, and switch it off instantly: when you revoke a link, connected viewers lose access right away. You also choose who gets in: anyone with the link, signed-in people only, or specific people, an organization, or a team. Only the session owner can change any of these settings.

Share-link passwords are never kept in plain text. They're encrypted at rest with AES-256-GCM and checked using a constant-time comparison, so the password itself stays protected even at the database level.

Yes. The Free plan runs live transcription and live translation with no credit card, so you can test Kalima on a real meeting first. When you need more minutes, sharing, or collaboration, you can move up to a paid plan.

There are five plans. Free, Plus, and Pro are for individuals; Team and Enterprise are for teams. Free opens with no card. Plus is 12.49 EUR/month, Pro is 21.99 EUR/month and our most popular tier, Team is 29.99 EUR/month per seat, and Enterprise is a custom plan you arrange with sales. Annual billing is available on every paid plan and lowers your effective monthly price.

Paid plans bill monthly or annually, and annual billing lowers your monthly price. You can upgrade, downgrade, or cancel anytime from your Billing dashboard. If you cancel, your plan runs through the end of the period you have already paid for. Prices are shown in EUR, with your local currency available at checkout.

Yes. Upgrades take effect right away with prorated billing charged at once. Downgrades apply at the end of your current period, so you keep everything until then. Before you confirm a change, you can preview the prorated charges.

We warn you before you reach your transcription minutes, uploads, or AI usage. Each of these has its own weekly limit. If you hit one, that feature pauses until your usage resets each week (Monday) or you upgrade for more capacity. Your transcripts stay safe either way.

You pay only for the people who have active access. Add or remove seats anytime and your billing adjusts automatically. Viewers who open a shared link are always free and never count as a seat.

Major credit and debit cards through Stripe. Enterprise customers can also pay by invoice on NET-30 terms. You can manage your cards and invoices anytime through the billing portal.

Yes. You can top up with pay-as-you-go credits anytime if you need more minutes or AI usage than your plan includes, without changing plans.

Yes. Open any session and generate a shareable link with its own web address. The link works the moment you create it, and you can add an optional password and an optional expiry date.

No. Anyone with the link can open the session right in their browser, with no paid seat or subscription needed. Viewers who follow a shared link are always free.

Viewers get a read-only view. They can read the transcript and play the audio, but they cannot record, create sessions, or change any settings. Each person's chat over the transcript stays private, so viewers never see your chat history.

Yes. You can protect a link with a password and switch it off at any time. Turning it off disconnects anyone currently watching, and expired links stop working. You can also scope access: anyone with the link, any logged-in user, a specific person by email, a whole organization, or a single team.

Yes. While you record with viewers connected, your transcript streams to them in real time, so they see the words appear as you speak, including the in-progress sentence. When a new viewer joins, they catch up to where you are, and you can see the live viewer count and whether the speaker is online.

Yes. Group related sessions into projects, which act as folders, and keep different clients or contexts apart with separate workspaces. Invite collaborators to a project by email with a role: viewer (read-only), editor (create sessions and edit settings), or admin (manage people and settings); the creator is the owner. You can change a person's role or remove them, and people can leave a project. Team plans are priced per seat, so you pay only for people with active access and billing adjusts as you add or remove seats.