ODIN Cortex
Please note, that ODIN Cortex is in beta and will launch in production end of Q2 2026. If you are interested in give it a test, please get in touch.
ODIN Cortex is the real-time intelligence layer for voice chat. It listens to your ODIN Voice rooms, turns speech into text as it happens, runs that text through AI-powered analysis and moderation, and streams the results to your backend, your dashboard, or serverless code you deploy in minutes — all without you running a single piece of audio infrastructure.
If your players, teammates or attendees are already talking in an ODIN room, Cortex is how you make that conversation searchable, moderatable, and actionable.
Point Cortex at an ODIN room and get back a live, speaker-attributed transcript — plus the tools to moderate it, summarize it, react to it, and build on top of it.
Why Cortex
A headless bot joins your ODIN room and transcribes every speaker separately — using OpenAI Whisper or a local engine — and posts each utterance back in real time.
Detect toxicity and hate speech on the transcript, then warn, mute or ban with a full, auditable moderation trail.
Deploy your own TypeScript that reacts to events — send an email summary, post to Discord, call your game backend — with no servers to manage.
Organize people before and during a call: lobbies with “ready up”, scheduled appointments with calendar invites, and drop-in open spaces.
What you can build
- Meeting & call summaries — auto-generate a recap, action items and a shareable transcript the moment a conversation ends.
- Voice moderation for games — flag toxic chat in ranked lobbies and apply mutes or bans that your game server enforces on the next join.
- Compliance & record-keeping — keep a durable, attributed record of voice conversations for support, dispute resolution or regulated industries.
- Live captions & accessibility — stream captions into your UI over Server-Sent Events as people speak.
- Voice-driven automation — trigger workflows (tickets, emails, webhooks) from what is said in a room.
How it fits into the 4Players Platform
Cortex is one of four products that share a single 4Players project. It sits on top of ODIN Voice (the real-time audio layer) and runs its serverless workloads on ODIN Fleet (global container hosting). ODIN Rooms — browser conferencing — is a typical source of the rooms Cortex transcribes.
You bring the rooms and the application logic; Cortex provides transcription, analysis, moderation, coordination and the event plumbing that ties it together.
For developers: the three ways in
Everything Cortex does is available programmatically. Pick the integration surface that fits your stack:
The official @4players/odin-cortex package — typed resources, live objects and Firestore-style real-time watch() subscriptions.
A clean, fully-documented REST API. Build a client in any language — every SDK is generated from this same spec.
Webhooks, Server-Sent Events and serverless function triggers — all driven by the same domain-event catalog.
Get started
The fastest path to your first transcript is the getting-started guide — create a project, grab an API key, install the SDK, and transcribe a room end to end.
Get StartedWhere to go next
- New to the platform? Read Concepts to understand sessions, gatherings, participants and authentication before you write code.
- Building an integration? Start with the TypeScript SDK or the REST API.
- Want examples? The Guides walk through real, runnable scenarios.
Cortex is currently in beta. The REST API and SDK surface described here are stable for building against; we’ll call out breaking changes clearly as the product moves toward general availability.