Make online voice chats safe for everyone

Request Demo
Monitor the health of your community and keep voice chats safe to improve member satisfaction and retention.

Safe for kids

What is acceptable between adults might not be the same with children. Look out for your underage community and track any suspicious or unsafe situations around them.

Detect kids and setup appropriate actions

Safe for women

Quickly detect situations of cross-gender harassment in your community based on your member’s voice gender.

Detect female voices to help prevent sexism

Safe for everyone

Map the patterns and behaviours of your community with their negative and positive expressions.

Use acoustic intelligence to set, improve, or calibrate your Safety & Trust policies.

Use conversational risk scoring to help decisions

Fun for everyone

Happy voices shape community engagement.
 Promote systemic change by rewarding good behaviour in the community with the appreciation they deserve.

Detect fun and engaged interactions in sessions

65% of online gamers have experienced “severe and/or sustained harassment”

Players who face abusive language or harassment on their first session are 320% more likely to never return.

How it works

Acoustic intelligence

SafeVox uses unique voice intelligence to capture tone and other markers such as laughter, screaming, and high emotions, among many other acoustic biomarkers. 
Make better predictions and spot problematic behaviour faster with a simple solution that scales with and adapts to the diversity of your community.

Extract

AI extracts information from voice in any language and labels acoustic events in the most relevant way.

Analyse

AI determines conversational risk based on key behaviours between speakers, and highlights when vulnerable speakers are present.

For richer analytics, keyword spotting or transcription is applied to targeted segments.

Review

Content moderators can get to an accurate decision faster when presented with summaries of acoustic intelligence and most important highlights on speaker activity.

Saving time on: 

  • interpreting information
  • playing back the audio
  • reading transcripts
Minimising:

  • effort to make a decision
  • exposure to disturbing content

Features

The market’s most powerful audio moderation tool that uniquely combines the acoustic (tone) and semantic (words) modalities for superior accuracy and automation capabilities.

“The problem was HOW you said it!”

Conversational risk

Context is everything, more so with cross-culture communities. Human vocal expressions are universal enough to give you an understanding of how people “feel” during social interactions.
Get a better sense on the impact of someone's behavior, regardless of their intent.

Use it to score the behaviours of speakers beyond the words they say.

“Well... That escalated quickly!”

Sentiment heatmap

Get an immediate sense of the speakers' emotional journey before even clicking "play". From hot to cold zones, you can detect hotspots in the conversation to guide your analysis.

Use it to navigate through the audio and quickly spot the incidents that occurred.

“Why do you speak to her like that bro? She's alright!”

Voice gender

Narrowing down intent and meaning behind words becomes exceptionally hard and nuanced in cross-gender interactions, more so when one is a relative minority in a community.

Use it to help you better determine conversational risk in cross-gender settings.

“Dude, you’re talking to a kid, chill!”

Underage speaker

Don’t rely on online social justice warriors to protect the vulnerable members of your community.

Use it to check up on underaged members and make sure you double down on their safety.

“What is this horrible noise? Mute!”

Audio disruption

Get real-time information when audio disruption impacts the quality of the conversation. 
Intentional or not, the presence of disruptive noise negatively affects the user experience.

Use it to enable actions such as mute, kick, or to update a speaker's “community score”.

Deployment

Active moderation

Near real-time analysis on recorded audio streams to support human moderation.

Post-assessment

Review difficult cases and fine-tune the system for continuous improvements.

Live monitoring

Real-time analysis on live audio streams to support 
automated moderation.