Tutorial
Connect usage tutorial for better live translation.
Set up your voice identity, language variants, conversation context, pronunciation dictionary and speaker labels so Connect has the right information before your next call starts.
This tutorial starts after Connect is installed. If you still need the basic audio setup for Windows, macOS or Linux, start with the main usage guide. Once the app opens correctly, the settings below help Connect translate more accurately, pronounce important words correctly and keep multi-speaker calls easier to follow.
Conversation flow: pick how translation moves
On the main Connect screen, choose the language you speak in the From selector and the language your listener should hear in the To selector. Add language hints when you expect to hear a specific language from other speakers. Then choose your microphone and speaker at the bottom of the window.
Translation modes
- Unidirectional: use this when you only need your voice translated outward to others. Your microphone is translated, but you hear the original audio from others.
- Bidirectional: use this when both sides of the conversation should be translated. Both your voice and incoming speech are translated.
Processing modes
- Streaming Mode: use this for natural, continuous translation flow. Best for free-flowing conversations where pauses are minimal.
- Instant Mode: use this when you prefer short, clean translations after natural pauses. Best for structured calls like presentations or interviews.
Enhancement features
- Speaker Labeling: turn this on for group calls so transcripts can separate who said what. Use it for interviews, team calls, demos and support calls with several people. Keep microphones separated when possible, because overlapping voices are harder to label. Combine it with language hints.
- Language Hints: when you know which language(s) other participants will speak, add these as hints. This helps Connect focus on these languages for better accuracy, while still detecting any other languages automatically.
- Emotion Transfer: when enabled, Connect preserves the emotional tone of your voice in the translated speech. Available on Pro plans.
- Noise Suppression: automatically reduces background noise from your microphone input for clearer translation quality.
Create your voiceprint
A voiceprint lets Connect keep your translated voice closer to your identity. Open the portal, go to Voice Library, then create a new voiceprint. Use a quiet room, speak normally and avoid long pauses. If the clip has clicks, silence or background noise, trim or re-record it before submitting.
- Name the voiceprint so you can recognize it later.
- Select the language used in the recording.
- Add a short voice description if it helps distinguish several profiles.
- Upload a clean audio clip or use Record Now.
- Click Create Voiceprint, then select it in Connect before your call.
Create a voice variant
Like voiceprints, voice variants let you create specific voice profiles for different use cases. Open Voice Library and create a new voice variant that fits the target language and tone of your meeting.
Best practice: test the variant with How I Sound? before a real call. You can catch volume, pacing or language mismatches without involving another person.
Add conversation context for specialized calls
Context tells Connect what the conversation is about before it starts. This is especially useful for medicine, legal meetings, sales calls, finance, technical interviews, product demos and any call with names or uncommon vocabulary.
- Open the portal and select Conversation Context.
- Enter a clear profile name, such as Medical Consultation, B2B or Legal Meeting.
- Click Create Profile.
- Add the terms, phrases and expected language pairs for that call type.
- Activate the profile before joining the meeting.
Build a pronunciation dictionary
The pronunciation dictionary fixes words that often sound wrong in generated speech: company names, people names, product names, acronyms, medication names, cities and technical terms. Add the written word on the left and the pronunciation Connect should say on the right.
| What to add | Example | Why it helps |
|---|---|---|
| Brand names | Cartesia, Xiaomi | Keeps product and company names recognizable. |
| Acronyms | MRR, SQL, API | Prevents letter-by-letter or wrong-word pronunciation. |
| Names | Nguyen, Saoirse | Makes introductions and support calls feel more natural. |
Recommended call workflow
For the cleanest result, prepare your profile before the call instead of changing settings while people are already speaking.
- Choose the correct source and target languages.
- Select the voiceprint or voice variant you want to use.
- Activate the conversation context for the meeting topic.
- Activate the pronunciation dictionary for names and special terms.
- Turn on speaker labeling for group conversations.
- Press How I Sound? for a quick private test.
- Press CONNECT-ON when you are ready to speak live.