Real-Time Video Intelligence API

See every frame.
Understand everything.

Object detection + language reasoning delivered as a single WebSocket stream. Built for production, designed for developers.

Free tier · No credit card · 10,000 requests/month

Live
import { ArviumClient } from '@arvium/sdk'

const arvium = new ArviumClient({
  apiKey: 'arv_live_...'
})

const stream = await arvium.analyze({
  source: 'https://stream.example.com/live.m3u8',
  pipeline: ['yolo', 'vlm', 'transcription'],
})

Trusted by engineering teams worldwide

TechCorpStreamLabsMediaAIVisionOSDataFlowCloudRun

The Pipeline

From stream to intelligence in milliseconds

Step 01

Connect

Send any video source: HLS stream, RTMP, file URL, or webcam.

POST /streams
{ "source": "rtmp://broadcast.example.com/live" }
Step 02

Process

YOLO detection + VLM language reasoning run in parallel on every shot change.

// Parallel pipeline
YOLO v11  → bounding boxes, labels, confidence
VLM/Qwen  → summary, activity, products, sentiment
Step 03

Stream

Receive structured JSON via WebSocket. Real-time, typed, reliable.

ws.on('context', (event) => {
  console.log(event.data.summary)
  console.log(event.data.activity)  // "product_demo"
  console.log(event.data.products)  // ["wireless earbuds"]
})

Capabilities

Everything intelligence requires

Real-time Object Detection

YOLO v11 at up to 60fps. Bounding boxes, labels, confidence scores per shot change.

Scene Language Reasoning

VLM analysis returns structured JSON: summary, activity type, products, sentiment, engagement cues.

Speech Transcription

Word-level timestamps via Deepgram or self-hosted Whisper. Transcript context feeds into VLM reasoning.

Semantic Video Search

Natural-language search across all recorded sessions. Dual visual+text embeddings in pgvector.

Shot Intelligence

Automatic camera cut detection using dual metrics. Process only what matters.

WebSocket Streaming

Persistent real-time connection. 6 event types. Typed payload schemas.

Built for engineering teams

[Placeholder testimonial]
Arvium replaced our entire custom video analysis pipeline. What took us 6 months to build, we replaced in a weekend.

Alex Chen

CTO, StreamTech

[Placeholder testimonial]
The WebSocket streaming API is incredibly well-designed. Typed events, low latency, and the VLM context is genuinely useful.

Sarah Kim

Lead Engineer, MediaAI

[Placeholder testimonial]
We process 50+ live streams simultaneously. Arvium handles the intelligence layer so we can focus on our product.

Marcus Johnson

VP Engineering, LiveComm

StreamTechMediaAILiveCommVisionLabsDataStreamCloudAI

Pricing

Start free, scale as you grow

MonthlyAnnualSave 20%

Free

For experiments and prototypes

$0/month
  • 10,000 requests/month
  • 1 concurrent stream
  • Community support
  • 7-day data retention
  • All event types
Most Popular

Pro

For production applications

$99/month
  • 500,000 requests/month
  • 10 concurrent streams
  • Email support
  • 90-day data retention
  • Semantic search
  • Priority processing

Enterprise

For teams with custom requirements

Custom
  • Unlimited requests
  • Unlimited streams
  • Dedicated support + SLA
  • Custom data retention
  • Private deployment
  • Custom model fine-tuning

Questions? Talk to us →

Ready to see everything?

Start analyzing video streams in minutes. Free tier includes 10,000 requests per month — no credit card required.