miKasa help center
For listeners, speakers and organisers.
Listening to a session
How do I join?
Scan the QR code shown in the room, or open the link and type the 6-character code. No account or app is needed.
How do I pick my language?
Tap the language button at the top left of the screen and choose from the list. You can switch at any time without losing your place.
Listen, Read or Both?
Listen plays the translation out loud. Read shows live captions. Both does the two at once, which is handy in a noisy room.
Can I use my own headphones or speaker?
Yes. Connect wired or Bluetooth headphones (or a speaker) to your phone the normal way before or during the session. The audio plays through whatever your phone is using.
I hear nothing
Check your phone is not on silent, raise the volume, and make sure you pressed Start listening (phones block audio until you tap). If the session says Paused or Starting soon, the speaker has not started yet.
Can I see the page itself in my language?
Yes. The buttons and help on this page follow the language you listen in. To pick a different one, tap the globe button.
How do I ask a question?
When the organizer opens questions, an Ask button appears at the bottom of the screen. Type your question in your own language and press Ask. Everyone sees it in their own language, and you can vote for other people's questions so the most popular ones get answered first.
My language is not listed
Use "Request a language" at the bottom of the language list. The organiser sees every request.
Can I get a summary afterwards?
If the organiser shares the minutes, a "Read the session summary" button appears on this page shortly after the session ends, in your language. Keep the page open for a minute or two while it is written.
The session is at capacity
Each organiser has a limit on how many devices can listen at once. Try again in a moment or let the organiser know.
Running a live session
How do I go live?
Open the session, press Go live, then turn the microphone on. Everything you say is transcribed and translated into every language you offered.
The speaker changes language mid-event
Tap the language under "Speaker is talking in" in the console. The microphone restarts in the new language and listeners keep their own language. Example: English speaker, then a Chinese speaker, with English, Spanish and French listeners.
Can it switch language on its own?
Yes. Turn on "Auto-detect" in the console (or in the session settings). When someone starts speaking another of the session's languages, the session switches to it and the console tells you. It needs AI provider speech recognition, because the browser's built-in recognition only listens for one language. Very short phrases such as "OK" or "Merci" do not trigger a switch.
Show the QR code
Press "QR screen" to open a full-screen page with the QR code and join code. Put it on the projector or a second screen.
Show live captions on the projector
Press "Captions screen" (on the event page or in the console) on a laptop connected to the projector or TV. Pick one language, or add a second one to show both. Use the + and − buttons for text size and the full-screen button. The QR code stays in the corner so people can follow on their phones; press the QR button to hide it. The screen does not count as a listener, so it never uses up your plan's listener limit, and the controls hide when the mouse is still.
Pause vs End
Pause stops translating but keeps listeners connected. End closes the session for everyone and saves the transcript.
Take questions from the audience
In the console, open Audience questions. Listeners can then ask from their phones in any of the session's languages and vote for each other's questions. You see every question translated into the speaker's language, with the most votes first. Mark one Answered, Hide anything inappropriate, or press Read to room to send it to every listener, translated.
What happens when I hide a question?
It disappears from every listener's screen straight away and can no longer be voted on. It is not deleted: it stays in your console list (press "Show hidden" to see it), where you can bring it back with Show again, and it is kept with the session for the organiser. Hiding does not tell the person who asked.
Does Read to room show the question or speak it?
Both, on listeners' devices. The question goes into the live feed like a typed announcement: it appears as a caption on every phone and captions screen in each person's language, listeners in Listen or Both mode hear it read out, and it is saved in the transcript. It is not played through the room's loudspeakers, so the speaker or moderator should still read it aloud for people without a phone. You need to be live to use it.
Is the speaker's audio recorded?
No. miKasa never stores audio. With cloud speech recognition, the microphone is sent to your AI provider a few seconds at a time to be turned into text, and each piece is thrown away once it is transcribed. With the browser's built-in recognition, the browser handles the audio itself (Chrome and Edge use their own online service). Only the text is kept: the original transcript and its translations.
Typing instead of speaking
Use the text box under the live feed to send an announcement. It is translated like speech.
Best microphone setup
Use a dedicated microphone close to the speaker, or feed the room mixer into the computer. Chrome or Edge give the best built-in speech recognition. Choose cloud speech recognition in the session settings for the highest accuracy.
Getting started
Set up in four steps
1. Add an AI key under AI Keys. 2. Create an event under Events. 3. Add a session and choose the speaker language and the languages to offer. 4. Open the console and go live.
What is an event vs a session?
An event is the conference or seminar. Sessions are the talks inside it. Each session has its own QR code, languages and transcript.
Do listeners need an account?
No. Only your team signs in. Listeners scan, pick a language and go.
Use miKasa in my own language
Press the globe button in the menu and pick a language. Menus, buttons and help are translated with your own AI key the first time, then remembered. Listeners' pages follow the language they listen in.
AI keys (bring your own key)
Which AI providers are supported?
For translation: Claude (Anthropic), OpenAI, Google Gemini, DeepSeek, Mistral, xAI Grok, Groq, Qwen, Together, DeepL and OpenRouter (one key for 300+ models). For speech recognition: OpenAI, Groq, Deepgram and ElevenLabs. For natural voices: OpenAI and ElevenLabs. You can also add any OpenAI-compatible endpoint, such as Ollama, vLLM, LM Studio or another router like OpenRouter, by pasting its address.
Who on my team can see our AI keys?
Everyone on your team can see the list of keys on AI Keys: the label, the provider, the first and last few characters, and this month's spend. Only the Owner and Admins can add, test, remove or set a budget on a key; Operators and Viewers cannot. Nobody can see a full key again once it is saved, not even the Owner: keys are encrypted and only the server uses them.
Should I use a DeepL key?
DeepL translates well, but it ignores your Glossary and the session notes, so names of people, products and acronyms can come out wrong. Use Claude, OpenAI, Gemini or another AI model when names and terms matter, or give the session a DeepL key only as the backup. DeepL also cannot write minutes.
Who pays for the AI?
You do, directly with the provider. miKasa never marks up AI usage. Keys are encrypted at rest and are only used for your sessions.
Which key does a session use?
Each session picks a translation provider and model. Mark one key as default to use it automatically. Speech recognition and voices can use a different provider.
How much will it cost?
Translation runs once per language, not once per listener, so 500 listeners in 3 languages cost the same as 3 listeners. When you set up a session you see an estimated cost per hour, and Analytics shows estimated spend by provider. Estimates use the providers' list prices; your provider's bill is the exact amount.
What if my AI key stops working mid-event?
Give the session a backup key. When you edit a session, pick a "Backup key" under Translation AI (and under Speech recognition if you use an AI provider for it). If the main key fails, is too slow or runs out of budget, the backup takes over straight away and listeners keep getting translations. The console tells the speaker, and Overview lists every switch for a week. The main key is tried again every few minutes and takes back over once it works. Use a key from a different provider for the best protection.
Set a spending limit
On AI Keys, press "Set a budget" under a key and enter the most to spend each month. Choose "Stop using this key" to pause translation when it runs out, or "Just warn me" to keep going. Speakers see a warning in the console at 80% and when it runs out. If a premium AI voice runs out, listeners hear their phone's voice instead. The count starts again on the 1st of each month.
Languages and voices
How many languages can I offer?
The limit on your plan is per session, not a total: each session can translate into up to that many languages (the speaker's own language does not count), and different sessions and events can use different languages. Over 50 languages are available, including Chinese, Spanish, French, Twi, Ewe, Ga, Hausa, Yoruba and Swahili.
Make names and terms come out right (Glossary)
Open Glossary in the menu, or the Glossary button on an event. Add names of people, your organization, products and acronyms. Tick "Keep it exactly as written" for names, or give the wording for each language (Suggest with AI fills it in for you to check). Terms can apply to all events or just one. Paste a list adds a whole speaker list in one go. Every live translation follows the glossary, and cloud speech recognition listens out for these words. DeepL keys ignore the glossary and the session notes, so use another provider when names matter.
Does the glossary apply to transcripts and minutes?
Yes. Translated transcripts are the live translations, so they follow it. The minutes, and their translations, are written with your glossary too, and so are translations of audience questions. The original-language transcript is what the speech recognition heard: cloud speech recognition is told your terms so it spells them right, but the browser's built-in recognition is not. If you fix the glossary after a session, press Write again on the minutes to redo them.
Device voice vs cloud voice
Device voice uses the listener's phone to read the translation aloud for free. Cloud voices (OpenAI or ElevenLabs) sound more natural and are included on every plan; you only pay your provider for what they read out.
Minutes after a session
What are minutes?
When a session ends, your AI key reads everything that was said and writes a summary, key points, decisions, action items and questions raised. They are then translated into every language of the session.
Where do I find them?
Open the event and press Minutes on the session. Pick a language to read it, then Copy or Download. Write again redoes them, for example after you fix the glossary.
Can listeners read them?
Only if you tick "Share with listeners". Listeners who stay on the page then see a "Read the session summary" button in their language.
Which AI key writes them?
The session's translation key, if it can write text (Claude, OpenAI, Gemini, DeepSeek and similar). If it is a DeepL key, another of your keys that can write is used. The cost is small, about the same as translating the session into one more language.
Team, analytics and billing
Invite colleagues
Go to Team and add people as Admin (everything except billing), Operator (runs events and the console) or Viewer (sees events and analytics). The Owner has full access including billing.
What does Analytics show?
Listeners over time, peak concurrent devices, minutes listened, the most popular languages, listen/read mode split, ratings, requested languages and AI usage.
How do I pay?
Open Plan & Billing, pick how to pay (card, mobile money or bank, depending on what is offered), then choose a plan. You finish on the payment provider's secure page and come straight back. Your plan switches on as soon as the payment is confirmed.
Paying by bank transfer or mobile money
Pick the bank or mobile money option, then choose a plan. You will see the account to pay into, the amount and a reference to put in the payment note. After paying, enter the transaction ID (and a photo of the receipt if you like) and press "I've paid". Your plan switches on as soon as our team approves it.
Talk to sales or support
Use Contact sales on Plan & Billing, or New message under Messages. Our replies appear under Messages, with a badge in the menu.
When do I renew?
Card subscriptions through Stripe renew automatically. Paystack, Flutterwave and mobile money payments cover one month or one year; Plan & Billing shows the paid-until date and a Renew button.
How does the listener limit work?
Your plan sets how many listeners can be connected at the same time, across all your live sessions. A listener is one phone or browser tab; listeners stay anonymous, so it is not tied to a named person, and when one leaves, someone else can join. Captions screens and your own team do not count. When you are full, new listeners are told the session is at capacity. Upgrade under Plan & Billing at any time.
One-off event? Use an event pass
An event pass gives you any plan's listener limit and features for a few days, paid once, with no subscription to cancel. On Plan & Billing pick Event pass, choose the day your event starts and the plan size. Your normal plan comes back by itself when the pass ends.
Transcripts
Every session keeps its original and translated transcript. Open the event, edit the session and use Download transcripts to get any language.
Rename or rebrand
Your organisation name, logo and accent color are under Settings.