Help Center

1. Go to tolk.fm/login. Sign in with your email and password.
2. New here? Ask your organisation's owner or admin to invite you, or use a magic link from the sign-in page to create your own workspace.
3. Add your name and company in Settings → Profile. Team invites live under Settings → Organization.

Listeners never need an account. They open the QR code or the join link and pick a language.

• Listener demo: hear recorded stages in any language at tolk.fm/try — no sign-in.
• Live demo: run a short live session from your own microphone at tolk.fm/demo.
• In your workspace: the dashboard has a Free demo · Try live translation card so you can test your mic and languages before a real event.

1. In the workspace, open Create. The first screen asks How will this event run?
2. Pick a setup type: Spotlight, Polylogue, Conference, or Studio. See Which setup type should I choose?
3. In Basics, name the event, set the spoken language and choose up to 13 target languages.
4. Conference events add a Configure step for stages, inputs and the translation engine.
5. In Share & review, check the summary and create the event. You land in its control room.

Titles, languages, engines, voices and most audio settings can be changed after creation, many of them during the broadcast. Spotlight and Conference events can switch type; Conference and Studio can switch both ways.

Every event has a listener link and QR code. Conference and Studio events also have a deep link and QR code per stage.

• Download the QR as PNG or SVG, or as a Branded poster (PDF) in a horizontal or vertical layout.
• Project it on the title slide and on break slides.
• Send the link in the programme, an email, or a messaging group.
• Share hub: collects the listener link, stage links and the speaker monitor link (when enabled) in one place.

Replace the code-based link with a readable one:

tolk.fm/my-conference-2026

1. Open Event settings and find Premium join link slug.
2. Enter 3–40 lowercase letters, numbers or hyphens. Slugs are unique — you will see a message if one is taken or reserved.
3. Save. The new link works at once and the code-based link keeps working.

Bridge is not an event type: it is a desktop app that translates your own voice in any call. See Bridge below.

• Best for: tours, workshops, classes, briefings and small sessions with one main speaker.
• Audience: up to 50 listeners on their own phones or laptops.
• Equipment: a laptop or phone microphone works; a headset microphone sounds better.
• Speaker link: the speaker can broadcast from their own phone with a speaker link, without signing in.

• Stages: up to 12 stages per event, each with its own inputs, languages and QR code. Listeners pick a stage, then a language.
• Inputs: a mixer feed through a USB audio interface, an RTMP stream from vMix, OBS or ATEM, or several devices combined. See Audio & inputs.
• Speaker monitors: stage speakers or moderators can follow the translation and manage the audience queue on their own device.
• Human interpreters: keep them for high-stakes sessions and use Tõlk for the rest.

Up to 10 people each speak their own language and hear everyone else in the language they chose. It works like an audio-only call with raise-hand signals.

• Each participant chooses a speak language and a listen language.
• Use it next to your video tool, or in person with headphones.
• Polylogue rooms are not recorded.
• A Polylogue event cannot be switched to another setup type later.

Studio turns a recording or a script into reviewed language versions. Studio events open on Cues Studio, where every line can be checked before anyone hears it.

1. Upload a video or audio file, or import it from a link.
2. Choose the program type: Prerecorded plays the recording itself (film, replay); Scripted live follows a performance in the room, cue by cue (theatre, ceremony).
3. Add context if you like — the subject, a synopsis, script, slides or glossary, and notes per language. Cues Studio reads the whole program first and builds a term sheet, so names and terms stay consistent from line to line.
4. Pick a recipe per language: Subtitles, Standard, Premium, or the speaker's own voice. See Studio prices.
5. Review the transcript and translations. Use Translate again or wording help on any line, edit lines, move or push them later, transcribe a stretch again, and assign voices per speaker. Cues Studio flags missed speech and voices to fix.
6. Export the transcript to Word and import your edits back; changed lines are voiced again automatically.
7. Publish: broadcast on a schedule on the Tõlk listener page, or download the language tracks and captions for your own player.

Studio is in pilot with selected customers. Some steps may run with our team — we tell you which. A Conference event can switch to Studio (and back) to add prepared parts such as a recorded keynote followed by live Q&A.

Bridge is a desktop app for macOS, Windows and Linux. It adds a translated virtual microphone you can pick in Zoom, Teams, Google Meet, WhatsApp, Discord, OBS and most games — you speak your language and they hear theirs.

• You speak / They hear: one target language per direction.
• What you hear: incoming translation of the call on macOS 14.2+, Windows 10 (2004+) and 11, and Linux (PulseAudio or PipeWire).
• Output: Translation only, Original + translation, or Bypass.
• Windows: beta builds may need a virtual audio cable; the tester guide explains it.

Installers go to beta testers only. Request Bridge beta access.

In Online meeting, Tõlk joins your Zoom, Microsoft Teams, Google Meet or ClickMeeting session and translates what is said for listeners on the Tõlk join page.

1. Meeting: paste the meeting link.
2. When: join now or at the scheduled start.
3. For listeners: optionally show the webinar video, add a YouTube live link, and show speaker names on captions.

Streaming platforms such as YouTube Live, Vimeo, Twitch, Restream and StreamYard connect through RTMP — see Integrations.

In Speakers & schedule, add sessions with a title, start and optional end time, stage and speakers. Turn on Show to listeners and listeners see On stage and Up next in their menu.

Speaker details also help the engines: per-speaker voices follow each speaker's pronouns, and caption speaker names come from here.

• Live translation is available for every setup type.
• Scheduled playback of prepared tracks runs from Studio events.
• Prepared + live (a recording followed by live Q&A) is available in Studio pilots.

In Audio settings, choose Simple / mobile guide for a single device or Advanced AV for several feeds. Advanced AV lets you mark a feed as General / room or Audience / stage, lock it to a device, and Combine multiple inputs.

• Use a directional or headset microphone close to the speaker rather than a room mic.
• Set gain so speech peaks around –12 dBFS. Clipping and very low levels both hurt accuracy.
• Prefer wired connections. Bluetooth microphones add delay.
• Speaker focus in Audio settings translates only the person on a clip-on or headset mic and leaves room noise out. Turn it off when people share one room mic.

1. Ask the sound engineer for an aux send or direct output from the desk.
2. Connect it to the control room laptop with a USB audio interface, or send an RTMP stream from your encoder.
3. Select that input in Audio settings before going live.
4. Keep the send level steady — riding the fader during the event makes translation quality uneven.

A clean desk feed usually beats a lavalier near the speaker, especially in large or echoing rooms.

Before the broadcast, the control room checks your inputs, languages, voice and monitor and marks each Ready or Needs setup. Checks include microphone access, a stable audio source, at least one target language and, for rehearsals, two distinct devices (one speaking, one listening).

Rehearse in the real venue with the real microphone whenever you can.

In the Audience participation card, turn on Raise hand from phones so listeners can ask to speak from the join page. Their speech is translated for everyone like the main speaker's.

• Listeners tap Raise hand to speak and join the queue. The organiser or a speaker monitor approves, denies or ends each turn.
• Skip the approval queue gives the floor straight away.
• Open mics at once lets up to 8 people speak at the same time.
• Audience mic for speakers lets speakers use a phone as a roaming audience mic.

A speaker monitor link shows a speaker or moderator the translation as it happens and lets them manage the raise-hand queue without the organiser dashboard. Set it up in the Speaker monitor card in Audio settings, choose whether it translates the general or the audience input, and share the link from the share hub.

Listeners can turn on the original audio under the translation. In Audio settings, Original audio for listeners sets whom it starts on for:

Live AI translation trails the speaker by a few seconds: the engine has to hear a phrase before it can translate it.

For dictionary-assisted languages, Wait for complete thoughts in Event settings trades a little delay for whole sentences. Tell your audience at the start that translation trails the speaker slightly.

Headphones keep translated audio out of the event microphone and stop phones from disturbing neighbours. Ask attendees to bring earphones or hand out earbuds at registration. Listeners without headphones can choose Captions only or Ear Speaker when they join.

Do not put the input device on shared guest Wi-Fi. If it drops, translation stops for every listener on that stage.

Each event runs up to 13 target languages at once. Which languages you can pick depends on the engine for that language.

Qwen can be used once the account owner allows processing outside the EEA in Settings. That setting applies to every event in the account and can be turned off at any time.

Every language runs in one of three modes:

• Per language: set Translation mode for new languages and Default direct engine for new languages in Event settings, then change any single language as needed.
• Automatic failover: if an engine fails during the broadcast, Tõlk moves that language to another engine and keeps going.
• Speech recognition: for captions and dictionary-assisted languages, choose OpenAI, ElevenLabs, Google Gemini or Microsoft Azure under Speech recognition.

Test your speaker and content with more than one engine before a big event.

• Speaker voices: in Event settings, give each speaker a matching voice. Voices follow the pronouns set in Speakers & schedule.
• Studio voices: OpenAI and ElevenLabs voices for Direct + studio voice and dictionary-assisted languages. Preview a voice before you choose it.
• Speech speed: in Audio settings, set one speed for all languages from 0.25× to 1.5× and override it per language. Changes apply live, from the next sentence.
• Gemini keeps the speaker's own voice and has no voice picker; Qwen has no voice or speed controls.

Live translation never clones a speaker's voice. In Studio, the speaker's own voice recipe uses ElevenLabs dubbing to keep each speaker's voice in the recording's translations.

Add entries in Dictionary so speech recognition and translation get them right. Each entry is a Term, Person, Organization, Place or Acronym.

• Also recognized as: other spellings or the way people mispronounce it.
• Written translation: how it should appear in each language.
• Read as (voice only): how the voice should say it.

Quality drops with:

• Several people talking at once
• Fast speech with few pauses
• Strong accents or dialects the models know less well
• Specialist jargon not in your Dictionary
• Music, applause or crowd noise in the feed

Do not rely on Tõlk.fm alone for medical, legal, financial, safety, emergency or other high-stakes decisions. Provide human interpreters where accuracy is critical or legally required.

1. Scan the QR code or open the join link.
2. Pick a stage (Conference and Studio events with several stages).
3. Pick a language, or search the list.
4. Choose how to listen: Headphones, Ear Speaker, Speaker, or Captions only.

In the player, listeners can change language, turn original audio on or off, open live captions, and see Speakers and schedule.

Listeners can read live captions on their phone, switch between original and translation, change font and text size, and jump back to live. A full-screen caption view is available for projecting captions in the room.

• The first summary appears after about 5 minutes, then a new one every 5 minutes, with key quotes.
• Listeners can ask for the recap by email. In Live captions, turn on Publish to join link to let them review the transcript and summaries after the event and to send the recap emails.
• Published recaps can be copied or downloaded as a transcript.

A short feedback card asks listeners to rate stream quality, translation quality and ease of use and leave a comment. Results appear in Analytics.

Turn on Record broadcast in Event settings. After the broadcast, Recordings holds the original audio and one audio file per translated language, plus the webinar video for online meetings. Recordings are private to your organisation and never shown to listeners.

With Cost saver warm-up, languages nobody listened to are not translated, so they are missing from the recording. Use Always ready when you need every language recorded.

Live captions keeps the captions archive for each broadcast. Download it as JSON, Markdown, SRT, VTT or plain text, and publish it to the join link if you want listeners to see it.

Analytics shows listeners now, peak and unique listeners, time listened, history by stage, the rating breakdown and recent comments.

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

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.

The same rates apply when you re-voice a program after editing it. Failed jobs refund what was reserved for them.

• 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.
• With Cost saver warm-up, a language is only translated while someone listens to it or a caption screen shows it.

Estimate ≈ hours × target languages × starting price

1. Tell us about your event on the pilot form, or email hello@tolk.fm.
2. Include the date, duration, number of stages and languages, and how the audio reaches us.
3. The Tõlk team confirms your quote and how you pay before the event.

• For invoices, payment and billing questions, email billing@tolk.fm with your event ID.
• Refunds are given where the law requires or where we agree case by case. AI provider costs already used may not be refundable.
• EU consumers have a 14-day withdrawal right for distance contracts, which may not apply once a digital service has been fully performed.

In Settings → Organization, use Invite member and pick a role: Member, Admin, Billing or Owner. The person accepts the invite from the email link.

Create API keys in Developers to create events, manage language channels, launch and stop sessions, and set up RTMP ingest, RTMP pushes and direct stream links from your own platform. Read the API docs.

To have data deleted sooner, contact privacy@tolk.fm. See also the Privacy Policy — Retention.

Audio and text go only to the providers your event uses. We use business API terms with data-processing agreements, and your data is not used to train their models.

Full details are in the Privacy Policy — AI providers.

• Join code, event and stage
• Chosen language and listening mode
• Playback session status and connection logs
• Device and browser type, IP address and approximate location
• Feedback, raise-hand requests and recap email address — only if the listener gives them

Listener data is not sold. It is shared only with the infrastructure providers that run the service.

• Tell attendees that audio is sent to AI providers for live translation — at the start and in writing (programme or signage).
• Tell them if you record the broadcast or publish the transcript.
• Have a legal basis for streaming audience voices, especially with raise hand and open mics.
• Do not capture children's voices without documented parental consent.
• If you allow Qwen, assess whether sending attendee voices to Alibaba Cloud outside the EEA is lawful for your event.
• Disclose AI-generated content where your jurisdiction requires it, and provide human interpreters where rules demand them.

Laws differ by country and industry. Ask legal counsel for regulated events.

1. Check that the input device is online and the audio meter moves in Audio settings.
2. Check the language: with Cost saver warm-up, a language only runs while someone listens.
3. Change the affected language's engine or translation mode in Event settings.
4. Still stuck? Email hello@tolk.fm with your event ID.

• Check the input device’s connection — round-trip times above 150 ms add delay.
• Try another engine for the slow language.
• For dictionary-assisted languages, turn off Wait for complete thoughts if you need the shortest delay.
• Keep the feed steady and well paced; long silences and very slow speech make engines wait longer.

• Chrome: lock icon → Site settings → Microphone → Allow, then reload.
• Firefox: lock icon → clear permissions, reload and allow again.
• Safari: Settings → Websites → Microphone → tolk.fm → Allow.
• iOS: Settings → Safari → Camera & Microphone Access → allow.

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

• Check they picked the right stage and language.
• Ask them to tap play again — phones block sound until the page is tapped.
• Check the phone is not on silent and the volume is up, or suggest Captions only.

The event ID is in the control room URL:

tolk.fm/events/[event-id]

Include it in support emails so we can find your event quickly.