Voice AI Tips & Troubleshooting
Solutions to common audio, transcript, and connection problems, and how to set up your environment for the best results.
Most SpeakNow issues fall into one of three categories: audio quality, connection stability, and score interpretation. This guide addresses the most common problems and the steps to resolve them.
Common issues and solutions
Audio lag or brief silence (1–3 seconds)
What happens: The tutor pauses for longer than expected, or your words appear in the transcript with a short delay.
Why it happens: Voice AI processes speech in chunks. A 1–3 second delay is normal on a standard broadband connection.
What to do: Press on and continue speaking normally. The session will catch up. Only disconnect if silence persists for more than 10 seconds.
Your speech appears duplicated in the transcript
What happens: A sentence you said appears twice in the conversation panel.
Why it happens: Your microphone is picking up the AI tutor's voice through your speakers, and processing it as a second input (acoustic echo).
What to do: Switch to headphones or earbuds. If the problem persists, lower your speaker volume slightly so the microphone cannot pick it up.
The tutor cuts you off mid-sentence
What happens: The AI interrupts you before you have finished answering.
Why it happens: Voice AI uses silence detection to decide when you have finished. A very long pause mid-sentence, or a rising intonation that sounds like a question, can trigger the end-of-turn signal.
What to do: Finish sentences with a clear falling intonation and a deliberate pause at the end. Alternatively, switch to Push-to-Talk mode (PTT button in the toolbar) so the AI only listens when you hold the button.
Session disconnects during a practice lesson
What happens: The session drops and you see the Disconnected status.
What to do: Press Connect again. The session will resume and your progress up to the disconnect point is usually saved. A single disconnection is normal.
If disconnections happen more than twice in one lesson, check your internet connection stability before continuing.
Session disconnects during an assessment
What happens: The connection drops during the CEFR placement or IELTS mock assessment.
What to do: Reconnect and continue from where you left off — the AI will pick up the conversational context. If disconnected during IELTS Part 2, briefly summarise what you had covered before continuing your monologue.
Only restart the full assessment if session data was completely lost. This is rare.
Multiple reconnections in one session
What happens: You disconnect and reconnect three or more times in a single session.
What to do: Keep pressing on and reconnecting — completing a session with reconnects gives more useful feedback than abandoning it. If you exceed 3–4 reconnections, note the time and contact support with your session details.
Microphone not detected
What happens: The browser shows no microphone, or the Connect button is disabled.
What to do:
- Click the padlock icon in the browser's address bar.
- Find the Microphone permission and set it to Allow.
- Refresh the page.
On mobile: Settings → Privacy → Microphone → [Your Browser] → Allow.
Score or band seems unexpectedly low or high
What happens: Your CEFR level or IELTS band estimate does not match your expectation based on other assessments or your own perception.
Why it happens: AI estimates are affected by audio quality, background noise, and session length. A noisy room, an echo, or a session that ended early can all shift the estimate.
What to do: Ensure your next session is in a quiet room with headphones and runs for at least the minimum recommended duration. Results across three or more consistent sessions are far more reliable than any single data point.
Best environment setup
The right physical setup removes the most common causes of issues before they occur.
| Element | Recommendation |
|---|---|
| Device | Laptop or desktop for best built-in microphone quality |
| Headphones | Wired headphones or good-quality wireless earbuds to eliminate echo |
| Room | Soft furnishings (carpet, curtains) absorb echo; hard surfaces amplify it |
| Background noise | Close windows during high-traffic hours; mute notifications |
| Browser | Chrome or Edge for best WebRTC performance |
| Connection | Use Wi-Fi or wired Ethernet; avoid mobile data for assessments |
If you are on a restricted network (corporate VPN, school Wi-Fi), the network firewall may block WebRTC traffic. Try switching to a mobile hotspot for a test session.