Frequently asked questions

Tõlk.fm is live AI translation for events, meetings and recordings. Connect the speaker’s audio, choose the languages, and share a QR code. Listeners scan it, pick a language and hear the translation on their own phone — no app, no account, no receivers.

Spotlight for one speaker and up to 50 listeners; Conference for venues with stages, mixer feeds and large audiences; Polylogue (beta) for conversations where up to 10 people each speak their own language; Studio (pilot) for reviewed translations of recordings; and Bridge (desktop beta), which translates your own voice in any call or game.

No. Organisers work in the browser at tolk.fm and listeners open the join link in any modern browser. Only Bridge, the desktop app for calls, is installed — and it is in beta for testers.

Sign in, open Create, and pick a setup type: Spotlight, Polylogue, Conference or Studio. In Basics, name the event, set the spoken language and choose up to 13 target languages. Conference adds a Configure step for stages and inputs. Review and create — you get a listener link and QR code straight away.

• Listener demo: hear recorded stages in any language at tolk.fm/try.
• Live demo: run a short live session from your own microphone at tolk.fm/demo.
• In your workspace: use the Free demo · Try live translation card in the dashboard.

Before a real event, rehearse in the same room with the same microphone or mixer. Microphone placement and room acoustics strongly affect translation quality.

Conference has no fixed listener cap: Tõlk translates once per language and streams that to everyone listening to it. Spotlight is for groups of up to 50 listeners. More listeners never increase the price.

Spotlight: one speaker, one link, own microphone. Conference: several stages, venue audio, speaker monitors and a QR code per stage. Polylogue: a multilingual conversation of up to 10 speakers. Studio: a recording or script translated, reviewed in Cues Studio, then broadcast on schedule or exported.

Yes. In the Online meeting section, Tõlk joins your Zoom, Microsoft Teams, Google Meet or ClickMeeting session and translates it for listeners on the Tõlk join page. You can also show the webinar video and speaker names on captions.

Yes, with Studio (in pilot). Upload a file or import a link, add context such as a synopsis or glossary, and review every line in Cues Studio before publishing. Then broadcast it on schedule on the Tõlk listener page, or download the language tracks and captions.

Yes. Turn on Raise hand from phones in Audio settings. Listeners raise a hand from the join page, and the organiser or a speaker monitor gives them the floor. Their speech is translated for everyone. Up to 8 audience mics can be open at once.

A link for the speaker or moderator that shows the translation as it happens and lets them approve, deny or end audience turns without the organiser dashboard.

Yes. Languages, engines, voices, speech speed and most audio settings can be changed while the broadcast runs. Spotlight and Conference events can switch setup type, and Conference and Studio can switch both ways; Polylogue cannot switch.

Each event runs up to 13 target languages. The languages you can pick depend on the engine you choose for each one.

Qwen can be used once the account owner allows processing outside the EEA in Settings.

All three translate speech to speech in real time. They differ in coverage and voice:

You choose the engine per language and can change it during the event. If an engine fails, Tõlk moves that language to another one automatically.

Up to 13 target languages per event. You choose the engine and translation mode for each language separately, and Tõlk fails over to another engine automatically if one stops working.

Yes. Give each speaker a matching voice, choose OpenAI or ElevenLabs studio voices, and set speech speed for all languages or per language. Gemini keeps the speaker’s own voice. In Studio, the speaker’s own voice recipe keeps each speaker’s voice in the translation.

Add them to the Dictionary as a Term, Person, Organization, Place or Acronym, with other ways they may be heard, the written translation, and how the voice should say them. Dictionary-assisted translation uses every entry.

Translation accuracy depends on audio quality, speaking pace, accents, technical vocabulary, and background noise. For most conference and presentation contexts, AI translation is clear and understandable. However, AI outputs can contain omissions, mistranslations, or unsuitable phrasing.

Do not use Tõlk.fm as the sole translation source for medical, legal, financial, safety, emergency, or other high-stakes decisions. Always provide human alternatives where accuracy is critical or legally required.

Use the control room device’s microphone or a USB audio interface, add another device with a remote input link, let the speaker broadcast from their phone, send an RTMP stream from vMix, OBS or ATEM, or connect an online meeting. For venues, a direct feed from the mixing desk works best.

A few seconds. The engine must hear a phrase before it can translate it. Tell your audience at the start that translation trails the speaker slightly.

Engines follow one voice best. Give each speaker their own microphone, use Speaker focus for clip-on or headset mics, and use raise hand so audience members speak one at a time.

Yes. Listeners can turn on the original audio under the translation. In Audio settings, choose whether it starts on for nobody, for listeners of the stage language, or for everyone.

They scan the QR code or open the join link, pick a stage if there are several, choose a language, and choose how to listen: headphones, ear speaker, speaker, or captions only. No account or app is needed.

Yes. In Event settings, set a Premium join link slug of 3–40 lowercase letters, numbers or hyphens — for example tolk.fm/my-conference. Slugs are unique, and the code-based link keeps working.

Headphones are strongly recommended so translated audio does not leak into the event microphone. Listeners without headphones can use the ear speaker or captions only.

Yes. Listeners can read live captions in the original or their language. AI Summary adds a short summary every 5 minutes with key quotes, and listeners can ask for the recap by email.

The player reconnects automatically. Reopening the join link brings them back to the live stream.

Each product has one starting price, per hour per language. Listeners are always free.

Starting prices, excl. VAT. Your quote is confirmed by the Tõlk team. Get a quote.

Studio is priced by the recipe you choose for each language, per hour of the recording per language, excluding VAT. Cues Studio shows the estimate before anything runs.

One target language running for one hour is one billable hour; two languages for one hour are two. The number of listeners never changes the price.

Send the pilot form or email hello@tolk.fm with the date, duration, number of stages and languages, and how the audio reaches us. The Tõlk team confirms your quote and how you pay before the event.

Refunds are given where the law requires or where we agree case by case. Email billing@tolk.fm with your event ID.

Live audio is processed in real time and not stored unless you turn on Record broadcast. Recordings are private to your organisation. Captions and summaries stay in the captions archive and reach listeners after the event only if you publish them. Studio uploads are kept until you delete them.

Only the providers your event uses: OpenAI, Google Gemini, ElevenLabs, Microsoft Azure (speech recognition), and Alibaba Cloud for Qwen if the account owner allows processing outside the EEA. We use business API terms under which your data is not used to train their models.

No. Listeners join without registering. Session and connection logs are collected to run the stream and fix problems; feedback and a recap email address only if they choose to give them.

Tõlk.fm is operated by Pixibit OÜ, an Estonian company subject to GDPR. Organisers are responsible for using it lawfully at their events.

We recommend informing attendees:

• Audio from this event is being streamed to AI services to produce real-time translation.
• Translation is generated by an automated system and may not be perfectly accurate.
• Whether the broadcast is recorded or the transcript will be published.
• Audience members who raise a hand to speak are also translated and broadcast.

Local laws in some jurisdictions require advance disclosure or consent before recording or streaming audience audio. Check your local requirements before the event.

Yes. The Tõlk.fm Public API is a REST interface to create events, manage language channels, launch and stop live sessions, and get listener links, RTMP ingest, RTMP pushes and direct stream links. API keys are managed in the Developers dashboard. Documentation is at tolk.fm/developers/docs.

streamUrl is the public browser listener page for the live event (for example https://tolk.fm/your-code). For broadcast tools, use directStreamLinks from launch or POST /v1/events/{id}/direct-stream-links — WebSocket PCM16, HLS/AAC, or RTMP pull URLs on the media worker.

Yes. Send program audio to Tõlk.fm over RTMP from OBS, vMix, Wirecast or ATEM, share the listener link, and push translated audio per language over RTMP or pull it as HLS. YouTube and Vimeo do not accept third-party translation tracks natively, so run translated streams in parallel.

1. Check the input device is online and its meter moves in Audio settings.
2. With Cost saver warm-up, a language only runs while someone listens to it or a caption screen shows it.
3. Change the affected language's engine or translation mode in Event settings.
4. Refresh the control room page.

If the problem persists, email hello@tolk.fm with the event ID shown in the control room URL.

Audio instability is usually caused by one of:

• Weak listener Wi-Fi or mobile data — advise listeners to move closer to the Wi-Fi access point.
• The input device's connection — the device sending audio needs a stable connection to Tõlk.fm. Use wired Ethernet where possible.
• Too many devices on the same Wi-Fi — ensure the event Wi-Fi is separate from the public guest network.

A few seconds of latency is normal for live AI translation. Very high latency (more than 10–15 seconds) may indicate high network latency, overloaded AI provider capacity, or an unusually slow audio stream.

• High network latency between the organiser's device and our media worker
• Overloaded AI provider capacity (rare)
• An unusually slow or noisy audio stream requiring more buffering

Try another engine for the slow language. If the delay is consistently above 10 seconds, contact support with the event ID.

Make sure you have granted microphone permissions to the Tõlk.fm site in your browser. If you accidentally denied permission:

• In Chrome: click the lock icon in the address bar → Microphone → Allow.
• In Firefox: click the lock icon → Clear permissions → Reload and try again.
• In Safari: open Settings → Websites → Microphone → allow tolk.fm.

Using a USB audio interface? Pick it as the input in Audio settings, not the built-in microphone.

Live broadcasts need enough balance on your account to cover the event. Email billing@tolk.fm or ask for a quote, and the Tõlk team will set up your account for the event.

The event ID appears in the URL of the event control room page: tolk.fm/events/[event-id].