Random voice chat

In 2017, Juliana Schroeder, Michael Kardas and Nicholas Epley ran an experiment with a simple twist. People heard or read the same words - recruiters evaluating job candidates, for instance, or people listening to someone explain their views on a contested topic. When they heard the person's voice, they judged them as more thoughtful, more competent and more intelligent than when they read the identical words as text. The paper's title says it plainly: "The Humanizing Voice: Speech Reveals, and Text Conceals, a More Thoughtful Mind" (Psychological Science, 2017).

It is not hard to see why. Text strips out everything except the words. A voice carries hesitation, warmth, a laugh halfway through a sentence, the pause before something difficult. In text, a stranger is a username and a handful of lines. In a voice, they are obviously a person.

The obvious assumption is that more channels mean more understanding: voice is good, voice plus a face must be better. The psychologist Michael Kraus tested that in a series of studies published in American Psychologist in 2017. People interacted under different conditions and then tried to judge what their partner had been feeling. Those in voice-only conditions read other people's emotions more accurately than those who could both see and hear them.

The likeliest explanation is attention. Without a face to watch - or to perform for - people concentrate on what they hear, and the voice is where a great deal of emotional information lives. Faces are also easier to control than voices. We can hold a smile; it is much harder to keep tiredness or delight out of how we sound.

That is the main reason TalkLive has no camera. The other reasons are practical: no room to tidy, no appearance to be judged on, a fraction of the data, and no route for the kind of misuse that dogged random video chat. Voice chat vs video chat sets out the full comparison, including where video genuinely wins.

You press Tap to Talk, allow the microphone once, and join the live queue. When another adult is searching and your settings are compatible, the call connects - usually within seconds at busy times, a little longer in the quiet hours. You hear each other immediately. There is no ringing, no profile to look at, just "hi".

Optional preferences - country, interests - can steer the search, but they depend on who is online, and nobody's identity, age or location is verified. Most people find that global matching gives the most interesting calls anyway. TalkLive is busiest between about 15:00 and 21:00 UTC, measured from our own match counts; if you would like the local version, the country guides for India, the UK and the US convert it.

Calls use WebRTC, the open standard behind most browser calling, with the Opus codec compressing speech so that it holds up on weak connections and costs little data. Where a direct route between the two browsers is blocked - common on mobile networks and office Wi-Fi - the audio is relayed through TalkLive's TURN server. Either way it is encrypted in transit. How TalkLive works draws the whole path.

TalkLive does not record or store call audio. Automated safety checks run on your own microphone during a call - a loudness check on every device, speech-to-text on desktop browsers - and keep a short record only when abuse is detected. The other person can still record on their own device, so speak as though what you say could be kept: no surname, address, workplace or social handles with someone you have just met. Every call has Report and Block, and blocked people are never matched with you again.