Walk into the room
before the room.

Background

Conjure flips the camera. Use Runway's avatars not to be you, but to simulate the people who shape your life — so you can walk into the room before you walk into it.

One engineer, no team, real Runway pipeline, real Anthropic orchestration, no fakes in the demo path.

Introduction

Every adult life turns on a handful of conversations: asking for a promotion, pitching investors, delivering hard feedback, defending a thesis. Rare for the speaker, routine for the listener. There is nowhere to practice.

Conjure is the first product where you can walk into a room with the specific people you're about to face. The room scales — anywhere from one focused conversation to a full three-person panel like the CEO promotion ask. Each avatar has their own face, voice, and competing agenda. The CEO pushes back, the manager protects budget, the head of people negotiates timeline. They argue with you. They argue with each other. You get better in real time.

Then you leave. And here is where Conjure does something no other product can: the room keeps talking after you leave it. One click and the same avatars debrief about you among themselves, disagreeing on specific moments, forming distinct impressions. The CEO thinks you weren't aggressive enough on Q4. The manager disagrees and points to the timeline question. You watch the conversation you were never supposed to hear — the one every human wonders about after every important meeting.

Conjure is built on Runway's full stack. The live session runs as concurrent Runway Characters Realtime Sessions (gwm1_avatars), orchestrated through LiveKit Agents with a server-side LLM coordinator, WebRTC mic routing for turn-taking, and a text-relay topic so each avatar reacts to what the others just said. Persona faces are generated with gen4_image and turned into custom Runway avatars. The Postmortem is rendered separately: a coordinator LLM scripts the debrief between the same personas, each turn rendered as a Runway avatar video and stitched into a single film. Plus Runway seedance2text-to-video for an optional 10-second cinematic shot when the room was developing something visual — a writers' room pitching a film, a creative team building a music video.

Conjure is browser-based, consumer-facing, no headset, no enterprise sales motion. The first demonstration that Runway's avatars are not just for representing yourself outwardly, but for simulating the people who shape your life. The market is anyone with a hard conversation ahead of them — millions of users at $19/mo, recurring API revenue from day one, a roadmap from the room to discovery to delegation to a daily briefing your avatar generates while you sleep.

Most products end when you stop talking. Conjure keeps going.