Near-real-time video infrastructure for LLMs. It turns prompts, documents, and AI answers into narrated, mathematically exact explainer videos you can call from an app or agent.

How is this different from whiteboard AI video?

Whiteboard tools optimize for fast drawn explainers. Vivacity optimizes for coordinated, mathematically exact diagrams. Equation morphs, graphs, spatial proofs. When the visual has to be correct.

A plain prompt, a document, or an upstream model answer. Any topic works. Math and science are where precision shows most.

How fast and how cheap?

Near real-time on short clips. Careful short renders land around ₹7 / about $0.08. Designed so EdTech chatbots and agents can afford volume.

English, Hindi, and Hinglish narration today, with more language paths on the roadmap.

Yes. Submit a job, poll status, receive a video_url. Built so products and agents can embed generation without owning a render farm.

EdTech platforms, LLM apps, creator pipelines, and anyone replacing text dumps with motion that teaches.

Docs · v0.1

Introduction

Vivacity is near-real-time video infrastructure for LLMs. It turns a prompt, document, or model answer into a narrated, mathematically exact explainer video you can call from an app or agent.

It is not a slide deck tool. The output is a video URL. Math and science are where precision shows most.

Who it is for

Researchers who need to visualize a proof or derivation
Educators preparing concept explanations
Platforms integrating AI-generated educational content via API

Quick start

During early beta, access is invite-only. Request access from the team, then open the workspace with your beta code.

# conceptual flow
POST /v1/jobs  { "prompt": "Explain eigenvectors" }
GET  /v1/jobs/:id
→ { "status": "ready", "video_url": "..." }

Official SDKs and full auth docs ship with broader beta. Until then, use the workspace UI at /signin.

Pipeline architecture

Four stages run per request:

Prompt interpretation: extract mathematical intent and map it to a scene plan.
Code generation: turn the plan into executable motion code, with critique and repair.
Rendering: exact animation frames for the lesson.
Audio sync: align narration to animation keyframes.

REST API shape

Authentication will use API keys scoped to a workspace. Jobs are asynchronous. Poll status or subscribe to webhooks when available.

Authentication

Authorization: Bearer viv_live_xxx
Content-Type: application/json

Create a job

POST /v1/jobs
{
  "prompt": "Explain the chain rule",
  "resolution": "1080p"
}

→ { "id": "job_…", "status": "queued" }

Poll status

GET /v1/jobs/:id
→ { "status": "ready", "video_url": "https://…" }