Today we are partnering with 4Players, and it comes with a first for us. Amanda’s moderation now reaches voice chat. Until now Amanda has worked on text and images. Through ODIN, it can help keep voice rooms safe too, reading from a live transcript rather than touching the raw audio.
Voice is where modern communities come alive, and it has been the surface that is hardest to keep safe.
Voice chat is the hardest room to keep safe
Text leaves a record. A slur sits in a chat log, plain to find and easy to flag. Voice behaves differently. It moves fast, it disappears the second it is spoken, and the same sentence can be a joke between friends or a threat to a stranger depending on context.
That gap has left many voice rooms effectively unmoderated. Studios know harm happens there. Players who get harassed in voice tend to leave quietly and never say why. The cost surfaces later as churn, poor reviews, and a community that slowly turns cold.
ODIN brings players together, Amanda watches their back
4Players build ODIN, the voice infrastructure that connects players, learners, and communities across platforms. We build Amanda, the platform that keeps those spaces healthy. This partnership lets any ODIN customer plug Amanda into their voice rooms.
The fit is clean. ODIN Cortex, the intelligence layer behind ODIN, captures speech from active rooms and transcribes it. Amanda reads that transcript and considers the conversation around each line, so a pattern of harm shows up even when one remark looks innocent. It flags trouble within seconds. Aiba never handles the raw audio. The findings flow back to developers through ODIN Cortex, who decide what happens next: a warning, a mute, a ban, or a referral to support.
Built to catch what matters, fast
Amanda is tuned for the behavior that does real damage. It flags behavior like toxicity, grooming, and radicalization in near real time, fast enough to step in before a situation escalates.
It reads context, so harm that hides across a conversation still gets caught. When a case is too close to call, it goes to a human moderator for review, because some judgments still need a person. We have written before about why that balance matters here.
The payoff for studios
Safe rooms keep players around. The communities people return to are the ones that feel good to be in, and the ones that quietly empty out are almost always the ones nobody was watching.
For a studio, that makes moderation part of how you keep players, not a cost to cut. The partnership also keeps your team in charge. Amanda surfaces what it finds and connects to your existing flows, so the response is always your call. We make the broader case for this in trust pays off.
Privacy is part of the design
Voice data is sensitive, and we treat it that way. It is encrypted in transit and at rest, stored only when harmful speech is detected, and handled in line with GDPR. Your players keep their privacy and their safety at the same time.
Built on a decade of behavioral science
Amanda did not appear out of nowhere. It grew out of more than ten years of behavioral science research at the Norwegian University of Science and Technology (NTNU). That foundation shapes how we think about harm, and it is why Amanda focuses on patterns of behavior, where real harm shows itself.
Voice is the next real test for trust and safety, and we are glad to be taking it on with a partner who knows the terrain. If you run voice rooms and want to see how Amanda handles them, take a look at the Amanda platform or book a meeting.
