Built for Indian VFX & Media Studios

Your footage archive.
Finally searchable.

Niravi turns any video into a searchable knowledge base. Transcription, scene analysis, face detection, sound events, and AI chat — running on your own hardware. Nothing leaves your studio.

Schedule a Demo Try free on your Mac
Unreleased footage stays on your hardware
niravi — AI chat
you find all night exterior scenes with crowd VFX from the last three projects
────────────────────────────────
niravi Found 14 scenes across 3 projects
project_a · scene_042 · 02:18–03:44
project_a · scene_078 · 07:51–09:02
project_b · scene_019 · 00:44–01:58
+ 11 more …
────────────────────────────────
you what music is playing in scene_042?
niravi Music ID: "Raataan Lambiyan"
Timestamp: 02:34 – 03:11
Confidence: 97.4%
you
4×
faster than realtime
Rs.0
desktop, no limits
The Problem

Thousands of hours of footage.
None of it searchable.

Your team uses ChatGPT and Gemini for scripts and briefs. Your actual footage — the asset you've spent years building — is still locked in folders named COMP_v14_FINAL_FINAL.mov.

🎬

VFX studios

Finding a reference shot from 3 projects ago takes an afternoon of scrubbing. Your archive of award-winning composites, simulations, and plates is invisible to your team.

📺

Broadcasters & OTT

260,000+ hours of content and no way to find the right clip for a promo, recap, or compilation without a junior editor spending two days looking.

🖥️

Post & outsourcing houses

Shift handoffs are verbal. QC supervisors scrub manually. Artists re-search for the right frame every single project. The knowledge in your shots is lost the moment delivery happens.

What Niravi Does

Full video intelligence.
On your hardware. In minutes.

Process any video and every frame becomes searchable, queryable, and exportable. All models run on-device via Apple Silicon — Neural Engine, MPS, MLX.

🎙️

Transcription

Full transcript, speaker-separated, with timestamps. Whisper large-v3-turbo on MLX.

included
🎞️

Scene Detection

Every cut and transition detected with SOTA accuracy. TransNetV2.

included
🔍

Semantic Search

Find any scene by describing it in plain language. "Night chase with crowd." PE-AV 1024-dim embeddings.

included
👤

Face Detection & Tracking

Every face clustered and tracked across your entire library. Apple Vision + FaceNet.

included
🔊

Sound Events

303 categories of sound identified per frame — gunshots, crowd roar, wind, applause. Apple SoundAnalysis.

Neural Engine
🎵

Music ID

Exact track, artist, and timestamp for every song in the footage. ShazamKit.

On-device
🏃

Body Pose & Gesture

Full skeleton detection and gesture tracking across every person in frame. Apple Vision Neural Engine.

Neural Engine
💬

AI Chat Over Footage

Ask anything about your videos in natural language. Get timestamped, evidence-backed answers.

included
4 min
video processed
per
1 min
wall-clock time
M4 Mac Mini · 16GB · full pipeline
AI Integration

Already using Claude, ChatGPT,
or your own AI pipeline?

Niravi exposes an MCP server. Plug it into any LLM workflow and your AI assistant gains full access to your video archives — with no new UI to learn.

1

Install Niravi and ingest your library

Drop your footage in. Niravi indexes it automatically — transcription, scenes, faces, sound, embeddings. Takes about 15 minutes per hour of content on an M4.

2

Add one line to your AI config

One JSON entry in your Claude, Cursor, or custom agent config. The Niravi MCP server registers as a tool your AI can call.

3

Ask your AI about your footage

Your existing assistant — whatever LLM you already use — can now search, query, and reason over your entire video library.

What your AI can now answer
🔍 "Find every scene with crowd VFX we've done for Sony across all projects"
🎙️ "What does the lead say in Act 2 of the rough cut? Find the exact line."
🎵 "Which episodes have background music we haven't cleared for OTT rights?"
🎬 "Pull all night chase sequences from this season — I'm cutting a promo."
claude_desktop_config.json
{
  "mcpServers": {
    "niravi": {
      "command": "niravi-mcp",
      "args": ["--library", "/your/footage/path"]
    }
  }
}
Result
Claude can now search scenes, query transcripts, find faces, identify music, and answer questions about every video in your library.
Python — custom pipeline
import niravi

# connect to your indexed library
client = niravi.Client("/your/footage/path")

# semantic search
results = client.search(
  "night exterior crowd VFX",
  limit=10
)

# AI chat over footage
answer = client.chat(
  "Which scenes need music clearance?"
)
Result
Plug Niravi into any existing Python pipeline — Airflow DAGs, LangChain agents, custom dashboards, or your internal tools.
Who It's For

Built for studios that take
their footage seriously.

🎨

VFX & Post-Production Studios

"Your archive of award-winning shots is a competitive asset you can't search."

Find reference shots across all past projects instantly. Search composites, plates, and renders by what's in them — not by filename. Brief artists in seconds with exact timestamps and scene context.

📡

Broadcasters & OTT Platforms

"500,000 hours of content is only valuable if you can find what's in it."

Semantic search across your full library. Find clips for promos and recaps without a junior editor spending a week. Rights-aware — music identification flags every track automatically.

🏭

VFX Outsourcing & Service Studios

"Every shift handoff is a knowledge transfer problem."

Searchable shot history means the night shift knows exactly what the day shift did. QC supervisors review with full context. Revision cycles shorten when artists find the right frame in seconds.

⚙️

Pipeline & Technology Teams

"You're already building AI workflows. Here's the video layer."

REST API and MCP server. Plug into your existing LLM pipeline, internal tooling, or production management system. Fully scriptable — ingest via CLI, query via Python, build whatever you need on top.

Pricing

Simple pricing.
Per minute of video.

The same unit Twelve Labs uses — so you can compare directly. We're 7–15× cheaper because we run on Apple Silicon, not GPU cloud.

Desktop
Rs.0

Runs entirely on your Mac. Unlimited processing. No account needed.

  • All 8 capabilities included
  • Unlimited videos, no cap
  • Fully offline — no internet
  • MCP server included
  • Screenplay & shot list export
  • AI chat over footage
Download for Mac
Most Popular
Standard Cloud
Rs.5/min

Cloud processing for studios without Mac infrastructure. Pay per minute processed.

  • Everything in Desktop, plus:
  • Cloud processing — no Mac needed
  • REST API access
  • Priority queue
  • Team access & shared library
  • Standard support
Schedule a Demo
Pro Cloud
Rs.10/min

Higher throughput, dedicated capacity, and faster turnaround for production pipelines.

  • Everything in Standard, plus:
  • Dedicated processing capacity
  • Faster queue — no wait
  • Webhook & pipeline integrations
  • SLA guarantee
  • Priority support
Schedule a Demo
Enterprise
Talk to us

On-premise deployment, custom integration, volume pricing, and dedicated support.

  • On-premise on your hardware
  • Volume pricing negotiated
  • Custom API & MCP integration
  • Dedicated support engineer
  • NDA & data residency options
  • SSO, RBAC, audit logs
Talk to Sales
Example costs — 1-hour episode: Rs.300 (Standard)  ·  Full season (10 × 45 min): Rs.2,250  ·  10-hour archive ingest: Rs.3,000 Comparable international tools charge Rs.70+/min for fewer features.
Schedule a Demo

See it on your footage
in 15 minutes.

Bring a project — a rough cut, an archive reel, anything. We'll run it through Niravi live and show you what it finds. No setup required on your end.

Book a Demo Or try the desktop app free

Runs entirely on your hardware  ·  No data leaves your studio  ·  No account required for desktop