# Reactor

> Reactor is the developer platform for real-time AI — the missing infrastructure layer for real-time world models. We make real-time generative video accessible to developers, so they can build interactive AI experiences across media and entertainment, physical AI, and robotics.

World models are a shift in AI: from a tool you prompt into an experience you can step inside and interact with in real time. Reactor provides the platform — models, SDKs, and infrastructure — for building those experiences. Applications built on Reactor are already in developers' hands. Founded in 2026, Reactor emerged from stealth with $59M led by Lightspeed Venture Partners, with participation from WndrCo, NVentures (NVIDIA), Amplify Partners, and others.

- Website: https://www.reactor.inc
- Docs: https://docs.reactor.inc (full SDK reference + its own llms.txt at https://docs.reactor.inc/llms.txt)
- GitHub: https://github.com/reactor-team
- Contact: team@reactor.inc

Every page linked below also serves clean markdown: request it with the `Accept: text/markdown` header to get token-efficient text instead of rendered HTML.

## SDK Quickstart

Reactor ships a **JavaScript SDK** (React components/hooks + an imperative API) and a **Python SDK** (async, decorator-based) for connecting to AI video models running on Reactor's cloud.

The fastest way to start: scaffold a ready-to-run app with the `create-reactor-app` CLI — no manual setup.

```bash
npx create-reactor-app my-app --model=helios   # scaffold a starter app, then: cd my-app && npm run dev
```

Or add the SDK to an existing project:

```bash
npm install @reactor-team/js-sdk   # JavaScript / TypeScript
pip install reactor-sdk            # Python
```

Packages: [`@reactor-team/js-sdk`](https://www.npmjs.com/package/@reactor-team/js-sdk) on npm · [`reactor-sdk`](https://pypi.org/project/reactor-sdk/) on PyPI.

**Available models** — Reactor is an inference host; the "Built by" column credits the upstream lab.

| Model | Built by | Description |
|-------|----------|-------------|
| EdenLive | Tripo | Restyle games and live video with AI. |
| MiniMax FastH3 Omni | MiniMax + Hao AI Lab | Faster-than-realtime video generation with text, audio and image input. |
| Vidu S2-Avatar | Vidu | Talk live with a character made from one photo. |
| Vidu S2-Editing | Vidu | Edit a live video from one reference image. |
| H3 Reference Turbo Realtime | MiniMax | Video with synchronized audio, cast from one to nine ordered reference images. |
| MiniMax FastH3 | MiniMax + Hao AI Lab | Real-time video with synchronized audio, generated faster than you can watch. |
| Visko Orbis-Stable | Visko | Prompt-steerable video with realtime audio. |
| Visko Orbis-Dynamic | Visko | Realtime video that switches resolution mid-run, up to 4K. |
| LTX-Avatar | Lightricks | Bring a portrait to life — it speaks your script in real time, voice and lip-synced video generated together. |
| X2 | XMAX | Real-time streaming video-to-video editing. |
| Happy Oyster | Alibaba Cloud Model Studio (Bailian) | Describe a world and explore it live, generated frame by frame as you move. |
| LingBot World 2 | Ant Group | High-fidelity environments with real-time interactive output. |
| LingBot | Ant Group | High-fidelity environments with real-time interactive output. |
| Helios | ByteDance | Interactive real-time video generation with infinite streaming. |
| LongLive 2 | NVIDIA | Real-time multi-shot video generation with seamless shot changes and hard cuts. |
| SANA-Streaming | NVIDIA | Real-time video editing. |

**Connection lifecycle (both SDKs):** `DISCONNECTED → CONNECTING → WAITING → READY`. Wait for `"ready"` before sending commands.

**Authentication:** exchange your API key for a short-lived token server-side; never ship the key in client code.

```typescript
const r = await fetch("https://api.reactor.inc/tokens", {
  method: "POST",
  headers: {
    "Reactor-API-Key": process.env.REACTOR_API_KEY!,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    authorization_details: [
      {
        type: "session",
        resources: { models: { match: ["reactor/helios"] } },
        constraints: { max_sessions: 10 },
      },
    ],
  }),
});
const { jwt } = await r.json();
```

**React:**

```tsx
import { ReactorProvider, ReactorView } from "@reactor-team/js-sdk";

<ReactorProvider modelName="reactor/helios" jwtToken={jwt} connectOptions={{ autoConnect: true }}>
  <ReactorView />
</ReactorProvider>;
```

**Imperative:**

```typescript
import { Reactor } from "@reactor-team/js-sdk";

const reactor = new Reactor({ modelName: "reactor/helios" });
reactor.on("trackReceived", (name, track, stream) => {
  document.getElementById("video").srcObject = stream;
});
await reactor.connect(jwt);
await reactor.sendCommand("start", {});
```

For the complete API — every component, prop, hook, command, and the Python SDK — read the docs: https://docs.reactor.inc/llms.txt

## Models

Real-time world models hosted on Reactor's inference API. Each has a hosted, production API — no GPUs to manage.

- [EdenLive API](https://www.reactor.inc/models/eden-live) — Restyle games and live video with AI. Created by Tripo, hosted on Reactor. API context: https://www.reactor.inc/models/eden-live/llms.txt
- [MiniMax FastH3 Omni API](https://www.reactor.inc/models/fast-h3-omni) — Faster-than-realtime video generation with text, audio and image input. Created by MiniMax + Hao AI Lab, hosted on Reactor. API context: https://www.reactor.inc/models/fast-h3-omni/llms.txt
- [Vidu S2-Avatar API](https://www.reactor.inc/models/vidu-s2-avatar) — Talk live with a character made from one photo. Created by Vidu, hosted on Reactor. API context: https://www.reactor.inc/models/vidu-s2-avatar/llms.txt
- [Vidu S2-Editing API](https://www.reactor.inc/models/vidu-s2-editing) — Edit a live video from one reference image. Created by Vidu, hosted on Reactor. API context: https://www.reactor.inc/models/vidu-s2-editing/llms.txt
- [H3 Reference Turbo Realtime API](https://www.reactor.inc/models/h3-reference-turbo-realtime) — Video with synchronized audio, cast from one to nine ordered reference images. Created by MiniMax, hosted on Reactor. API context: https://www.reactor.inc/models/h3-reference-turbo-realtime/llms.txt
- [MiniMax FastH3 API](https://www.reactor.inc/models/fast-h3) — Real-time video with synchronized audio, generated faster than you can watch. Created by MiniMax + Hao AI Lab, hosted on Reactor. API context: https://www.reactor.inc/models/fast-h3/llms.txt
- [Visko Orbis-Stable API](https://www.reactor.inc/models/visko-orbis-stable) — Prompt-steerable video with realtime audio. Created by Visko, hosted on Reactor. API context: https://www.reactor.inc/models/visko-orbis-stable/llms.txt
- [Visko Orbis-Dynamic API](https://www.reactor.inc/models/visko-orbis-dynamic) — Realtime video that switches resolution mid-run, up to 4K. Created by Visko, hosted on Reactor. API context: https://www.reactor.inc/models/visko-orbis-dynamic/llms.txt
- [LTX-Avatar API](https://www.reactor.inc/models/ltx) — Bring a portrait to life — it speaks your script in real time, voice and lip-synced video generated together. Created by Lightricks, hosted on Reactor. API context: https://www.reactor.inc/models/ltx/llms.txt
- [X2 API](https://www.reactor.inc/models/x2) — Real-time streaming video-to-video editing. Created by XMAX, hosted on Reactor. API context: https://www.reactor.inc/models/x2/llms.txt
- [Happy Oyster API](https://www.reactor.inc/models/happy-oyster) — Describe a world and explore it live, generated frame by frame as you move. Created by Alibaba Cloud Model Studio (Bailian), hosted on Reactor. API context: https://www.reactor.inc/models/happy-oyster/llms.txt
- [LingBot World 2 API](https://www.reactor.inc/models/lingbot-world-2) — High-fidelity environments with real-time interactive output. Created by Ant Group, hosted on Reactor. API context: https://www.reactor.inc/models/lingbot-world-2/llms.txt
- [LingBot API](https://www.reactor.inc/models/lingbot) — High-fidelity environments with real-time interactive output. Created by Ant Group, hosted on Reactor. API context: https://www.reactor.inc/models/lingbot/llms.txt
- [Helios API](https://www.reactor.inc/models/helios) — Interactive real-time video generation with infinite streaming. Created by ByteDance, hosted on Reactor. API context: https://www.reactor.inc/models/helios/llms.txt
- [LongLive 2 API](https://www.reactor.inc/models/longlive-v2) — Real-time multi-shot video generation with seamless shot changes and hard cuts. Created by NVIDIA, hosted on Reactor. API context: https://www.reactor.inc/models/longlive-v2/llms.txt
- [SANA-Streaming API](https://www.reactor.inc/models/sana-streaming) — Real-time video editing. Created by NVIDIA, hosted on Reactor. API context: https://www.reactor.inc/models/sana-streaming/llms.txt

## Company

- [Reactor home](https://www.reactor.inc): What Reactor is and the platform overview.
- [Careers](https://www.reactor.inc/careers): Open positions and what we're building.

## Blog

- [From 49 Steps to Real Time: Inside FastH3-Omni](https://www.reactor.inc/blog/fast-h3-omni): Inside FastH3-Omni: the architecture, math and engineering behind real-time H3.
- [Optimizing the Wan VAE decoder for real-time world models](https://www.reactor.inc/blog/optimizing-wan-decoder): We cut the Wan VAE decoder's time per chunk from 133.8 ms to 60.2 ms, 2.2× faster, at the same output quality.
- [Reactor Expands Investors in Series A Round to Scale Real-Time World Model Infrastructure](https://www.reactor.inc/blog/nvidia-sapphire-ventures-join-reactor-series-a): Investment from NVIDIA and Sapphire Ventures comes as Reactor sees growing adoption across media, entertainment and physical AI
- [A kernel-centric path to real-time video generation on Trainium](https://www.reactor.inc/blog/reactor-on-trainium): Using the Neuron Kernel Interface, a Reactor–AWS collaboration tackled the dynamic shapes, memory access patterns, and cache management that make real-time autoregressive diffusion hard—building techniques that generalize across models.
- [Announcing Sol-H3 on Reactor: even faster MiniMax-H3, built by NVIDIA](https://www.reactor.inc/blog/fast-h3-sol-engine): Reactor partners with NVIDIA's SANA team to launch Sol-H3, powered by Sol-Engine: 5 seconds of 1344×768 video with native audio in 1.67 seconds on Blackwell, 3× faster than real time.
- [Visko Orbis 1.0 Launches on Reactor](https://www.reactor.inc/blog/visko-orbis-launch): Reactor partners with Visko to exclusively launch Orbis 1.0 — an infinite-length realtime video generation model, in Stable and Dynamic variants — on Reactor starting day zero.
- [LTX is Now Available on Reactor](https://www.reactor.inc/blog/ltx-on-reactor): Reactor partners with LTX to bring their newest open world video model — realtime, high-resolution, with native synchronized audio — to Reactor on day zero.
- [How to train a frontier-level world model](https://www.reactor.inc/blog/open-dreamer): We are releasing and open-sourcing the model and training code, and sharing all the lessons we learned along the way in this blog post.
- [Introducing Reactor: The Developer Platform for World Models](https://www.reactor.inc/blog/reactor-launch): Reactor emerges from stealth with $59M to build the platform for real-time World Models

## Legal

- [Privacy Policy](https://www.reactor.inc/legal/privacy)
- [Terms of Service](https://www.reactor.inc/legal/terms)

## Optional

- [Documentation](https://docs.reactor.inc): Full developer docs for the JavaScript and Python SDKs.
- [Docs llms.txt](https://docs.reactor.inc/llms.txt): Machine-readable SDK reference.