Skip to main content
ZeroTwo’s voice chat feature uses WebRTC and VAD (Voice Activity Detection) to enable real-time spoken conversations with AI models. Most voice issues are related to browser microphone permissions or network conditions.
Voice chat requires a supported browser (Chrome, Firefox, Safari, or Edge — latest versions) and an HTTPS connection. zerotwo.ai always uses HTTPS, so this is automatically satisfied.

​
Voice Troubleshooting

If clicking the microphone button in ZeroTwo does nothing:
  1. Check your browser. Voice chat is supported in Chrome, Firefox, Safari, and Edge on their latest stable versions. Older browsers or niche browsers may not support the required Web APIs.
  2. Ensure you’re on HTTPS. Microphone access is only permitted on secure origins. zerotwo.ai always uses HTTPS, so if you typed http:// manually, switch to https://.
  3. Check if another app is using the microphone. If a video call, recording app, or another browser tab is using your microphone, it may be locked. Close other apps using the mic and try again.
  4. Refresh the page. A stale page state can prevent the mic from initializing correctly.
  5. Restart your browser. Occasionally the browser’s media stack gets into a bad state and needs a fresh start.
If none of these work, check the browser permission settings in the next accordion.
If ZeroTwo never prompts you to allow microphone access, the permission was likely permanently denied at some point. Here’s how to reset it:Chrome:
  1. Click the lock icon (or info icon) in the address bar while on zerotwo.ai
  2. Click Site settings
  3. Find Microphone and change it from “Block” to “Allow”
  4. Refresh the page
Firefox:
  1. Click the lock icon in the address bar
  2. Click the arrow next to “Connection secure”
  3. Click More information → Permissions tab
  4. Find “Use the Microphone” and remove the “Block” permission
  5. Refresh the page
Safari:
  1. Go to Safari menu → Settings (Preferences) → Websites
  2. Click Microphone in the left sidebar
  3. Find zerotwo.ai and change the setting to Allow
  4. Refresh the page
Edge:
  1. Click the lock icon in the address bar
  2. Click Permissions for this site
  3. Set Microphone to “Allow”
  4. Refresh the page
If you can start a voice session but ZeroTwo doesn’t seem to hear you or respond:
  1. Wait 2–3 seconds after speaking. ZeroTwo uses VAD (Voice Activity Detection) to detect when you’ve finished speaking. There’s a brief delay before ZeroTwo starts processing. Don’t interrupt the session immediately — give it a moment.
  2. Check that your microphone is not muted in system settings. On Mac: System Settings → Sound → Input → ensure the input volume is not at zero. On Windows: Settings → System → Sound → Input device settings.
  3. Check the correct input device is selected. If you have multiple microphones (built-in + USB headset, etc.), make sure your browser is using the right one. In your browser’s microphone permission prompt, look for a dropdown to select the input device.
  4. Speak clearly in a quiet environment. Background noise can reduce VAD accuracy. Try again in a quieter space, speaking at a normal conversational volume.
  5. Try a different model. Some AI models may have slower response times that create the impression of no response. Switch to a faster model in the Model Picker.
Choppy audio and noticeable delay in voice sessions are typically caused by network issues:
  1. Switch to a faster or more stable network. Voice chat requires a stable connection. A minimum of ~1 Mbps is recommended, but lower latency matters more than raw speed. A wired ethernet connection is more stable than WiFi.
  2. Move closer to your WiFi router if you’re on WiFi, or switch to a 5 GHz band if your router supports it.
  3. Close bandwidth-heavy applications. Video streaming, large downloads, and other VoIP calls on the same network can cause audio degradation.
  4. Close other browser tabs that may be using audio or video resources.
  5. Try a different network entirely — if you’re on a corporate or school network with traffic shaping, try a mobile hotspot to test.
WebRTC-based voice sessions are sensitive to network jitter. Even if your download speed is high, network jitter (inconsistent packet timing) causes choppy audio.
Echo occurs when your microphone picks up the audio coming from your speakers, which is then processed and fed back to you.The fix is simple: use headphones or earbuds. This prevents the speaker output from reaching the microphone, eliminating echo entirely.If you must use speakers:
  • Lower the speaker volume
  • Increase the physical distance between your microphone and speakers
  • Use a directional microphone pointed away from your speakers
Voice sessions use WebRTC, which is sensitive to network changes:
  1. Stay on a stable network. Switching between WiFi networks, or between WiFi and cellular, will drop the WebRTC connection. If you’re on a mobile device, disable automatic network switching during voice sessions.
  2. Reconnect by clicking the mic icon again. Voice sessions can be restarted at any time — just click the microphone button to begin a new session.
  3. Check for VPN interference. Some VPN configurations interfere with WebRTC. Try disabling your VPN temporarily to test.
  4. Corporate network proxies can sometimes block WebRTC traffic. Try on a personal network or mobile hotspot to confirm.
  5. Check browser extensions. Some privacy-focused extensions block WebRTC. Try in an incognito window (extensions disabled) to test.
Voice session transcripts are saved to the chat after the session ends:
  1. Wait a moment after ending the session. Transcription processing happens server-side and may take a few seconds after you stop the voice session.
  2. End the session properly by clicking the stop/end button rather than just closing the tab. Abruptly closing the tab may prevent transcript saving.
  3. Ensure you’re logged in. If you started a voice session without being logged in, the transcript may not be saved. ZeroTwo saves transcripts to your account.
  4. Refresh the chat page if the transcript doesn’t appear within 30 seconds of ending the session.
If transcripts consistently don’t save, contact reed@zerotwo.ai.
ZeroTwo’s spoken and written response language follows your language setting:
  1. Go to Settings → Personalization → Language
  2. Select your preferred language from the dropdown
  3. This change takes effect immediately — the next voice session will use the selected language
Available languages: English, Spanish, French, German, Portuguese, Russian, Arabic, Hebrew, Hindi, Japanese, Chinese Simplified.
If you speak to ZeroTwo in a language different from your setting, ZeroTwo typically responds in the language you spoke in, regardless of the setting. The setting acts as a default when ZeroTwo can’t determine your preferred language from context.
See Supported Languages for the full language reference.

​
Voice Requirements Summary

RequirementDetails
Supported browsersChrome, Firefox, Safari, Edge (latest stable)
Connection protocolHTTPS (automatic on zerotwo.ai)
Minimum network speed~1 Mbps; low latency preferred
Microphone permissionMust be “Allow” for zerotwo.ai

​
Still Having Issues?

Email reed@zerotwo.ai with your browser, OS, network type, and a description of the voice issue.