What is an AI meeting notes API?

An AI meeting notes API is a programmable interface that transcribes meetings in real-time or from recordings, then uses AI to extract structured highlights such as action items, decisions, questions, and topic summaries. V100's AI meeting notes API uses Claude Haiku for highlight detection and supports 40+ languages with word-level timestamps and speaker diarization.

How does V100 extract action items from meetings?

V100 passes the full transcript through Claude Haiku, which identifies statements that constitute commitments, assignments, or next steps. Each action item is returned with the assigned speaker, the verbatim text, the timestamp where it was spoken, and a priority level. The AI distinguishes between hard commitments (e.g., 'I will send the report by Friday') and soft suggestions (e.g., 'we should think about redesigning that').

Does V100 support real-time AI meeting notes?

Yes. V100 supports both real-time meeting transcription via WebSocket with sub-second latency, and post-meeting analysis of uploaded recordings. Real-time mode delivers interim and final transcript events as participants speak, while post-meeting mode generates the full AI summary, action items, and highlights after the recording is processed.

AI MEETING NOTES API

AI Meeting Notes API
Automated Summaries, Action Items & Transcription

V100's AI meeting notes API transcribes meetings in real-time across 40+ languages, then uses Claude Haiku to extract structured highlights: action items with assignees, key decisions, open questions, topic summaries, and per-speaker sentiment. Every transcript includes word-level timestamps, speaker diarization, and a post-quantum digital signature (ML-DSA-65) that makes the record tamper-proof. Send a recording URL or stream live audio via WebSocket, and receive a complete meeting summary with exportable SRT, VTT, or JSON output in a single API call.

Get API Key — Free Tier View Integration Guide

40+ languages with auto-detection

Real-time transcription via WebSocket

Post-quantum signed (ML-DSA-65)

What the AI Meeting Notes API Does

Most meeting transcription APIs stop at speech-to-text. V100 goes further. After transcribing the audio, the API runs the full transcript through Claude Haiku to produce a structured analysis layer on top of the raw words. You get the transcript and the intelligence in a single response.

The API accepts a recording URL, a file upload, or a live WebSocket stream. It returns everything a human note-taker would capture: who said what, what was decided, what needs to happen next, and which topics were discussed. The difference is that the AI processes a one-hour meeting in under sixty seconds and never misses a detail.

POST /v1/meetings/analyze

curl -X POST https://api.v100.ai/v1/meetings/analyze \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "source": "https://storage.example.com/standup-2026-03-28.mp4",
    "language": "auto",
    "features": {
      "transcription": true,
      "diarization": true,
      "word_timestamps": true,
      "highlights": {
        "action_items": true,
        "decisions": true,
        "questions": true,
        "topics": true,
        "sentiment": true
      },
      "summary": true
    },
    "export": ["json", "srt", "vtt"]
  }'

Response — 200 OK

{
  "meeting_id": "mtg_4d7e9a2f",
  "status": "completed",
  "language": "en",
  "duration_seconds": 2340.6,
  "speakers": ["Sarah Chen", "David Park", "Maria Silva"],
  "summary": "The team reviewed Q1 metrics showing 23% growth in active users. Sarah proposed migrating the auth service to V100's SDK by April 15. David flagged a latency regression in the EU region and committed to a root-cause analysis by Thursday. Maria confirmed the design review for the new dashboard is scheduled for next Monday.",
  "highlights": {
    "action_items": [
      {
        "text": "Migrate auth service to V100 SDK",
        "assignee": "Sarah Chen",
        "due": "April 15",
        "priority": "high",
        "timestamp": 482.1
      },
      {
        "text": "Root-cause analysis on EU latency regression",
        "assignee": "David Park",
        "due": "Thursday",
        "priority": "high",
        "timestamp": 1204.7
      }
    ],
    "decisions": [
      {
        "text": "Approved migration to V100 SDK for auth service",
        "timestamp": 510.3
      }
    ],
    "questions": [
      {
        "text": "Can we get a budget estimate for the EU region expansion?",
        "speaker": "Maria Silva",
        "resolved": false,
        "timestamp": 1892.0
      }
    ],
    "topics": [
      { "name": "Q1 Metrics Review", "start": 12.4, "end": 420.8 },
      { "name": "Auth Service Migration", "start": 421.0, "end": 890.2 },
      { "name": "EU Latency Issue", "start": 891.5, "end": 1640.3 },
      { "name": "Dashboard Redesign", "start": 1641.0, "end": 2320.1 }
    ],
    "sentiment": {
      "overall": "positive",
      "by_speaker": {
        "Sarah Chen": "positive",
        "David Park": "concerned",
        "Maria Silva": "neutral"
      }
    }
  },
  "words": [
    {
      "text": "Alright",
      "start": 0.120,
      "end": 0.560,
      "confidence": 0.97,
      "speaker": "Sarah Chen"
    },
    // ... every word with timestamps, confidence, and speaker
  ],
  "exports": {
    "json": "https://api.v100.ai/v1/meetings/mtg_4d7e9a2f.json",
    "srt": "https://api.v100.ai/v1/meetings/mtg_4d7e9a2f.srt",
    "vtt": "https://api.v100.ai/v1/meetings/mtg_4d7e9a2f.vtt"
  },
  "pq_signature": "ml-dsa-65:0x3f8c1a..."
}

Real-Time vs Post-Meeting Analysis

V100 supports two modes for AI meeting notes. Use real-time transcription when participants need live captions and instant context, or post-meeting analysis when you need the most accurate summary from a recording. Both modes produce the same structured output; the difference is timing and accuracy trade-offs.

Real-Time Mode

Connect a live meeting via WebSocket. V100 delivers transcript events with sub-second latency as each participant speaks. Speaker identification improves throughout the session as the model learns voice signatures. Interim results arrive immediately; final, corrected results follow within two seconds. At the end of the session, the full AI summary, action items, and highlights are generated and delivered via webhook.

WebSocket streaming with sub-second latency

Interim + final transcript events

Live speaker diarization

End-of-session AI summary via webhook

Post-Meeting Mode

Upload a recording or provide a URL. V100 processes the file asynchronously with multiple passes for optimal accuracy. The speech model refines diarization and punctuation across the entire recording before Claude Haiku analyzes the complete transcript for highlights. Post-meeting mode produces the highest-quality summaries because the AI has full context from start to finish rather than processing incrementally.

URL or direct file upload

Multi-pass transcription for higher accuracy

Full-context AI analysis

Webhook or polling for results

AI-Detected Highlights

After transcription, V100 passes the full text through Claude Haiku to identify the five categories of meeting intelligence that matter most: action items, decisions, open questions, topic boundaries, and sentiment. Each highlight is returned with the exact timestamp and speaker attribution, so you can link directly to the relevant moment in the recording.

Action Items

Detects commitments, assignments, and next steps. Each action item includes the assignee (speaker), verbatim text, timestamp, due date (when mentioned), and a priority classification. The AI distinguishes hard commitments from casual suggestions so your task list stays clean.

Decisions

Identifies moments where the group reaches agreement or a leader makes a call. Returned with the decision text, the timestamp, and the participants involved. Useful for audit trails, compliance records, and making sure nothing decided in a meeting gets lost afterward.

Questions

Captures questions raised during the meeting and tracks whether they were answered. Unresolved questions are flagged separately so follow-up is straightforward. The API returns the question text, the speaker who asked, the timestamp, and a resolved/unresolved status.

Topic Segmentation

Automatically divides the meeting into named topics with start and end timestamps. Each topic is labeled based on what was discussed, creating a navigable table of contents for the recording. Users can jump to any topic by clicking its timestamp in your UI.

Sentiment Analysis

Reports overall meeting sentiment and per-speaker sentiment. Useful for managers tracking team morale, sales teams analyzing prospect calls, and customer success teams monitoring client satisfaction. Values include positive, neutral, concerned, and negative.

Executive Summary

A concise, natural-language summary of the entire meeting. Written in past tense with specific details -- names, numbers, dates -- rather than vague statements. Designed to be forwarded to stakeholders who missed the meeting without losing critical context.

Supported Languages

V100 transcribes meetings in 40+ languages with automatic language detection. AI highlights and summaries are generated in the same language as the transcript, or you can request English summaries regardless of the source language. Here are the top 20 languages by accuracy.

Language	Code	Real-Time	Post-Meeting	AI Highlights
English	en	Yes	Yes	Yes
Spanish	es	Yes	Yes	Yes
French	fr	Yes	Yes	Yes
German	de	Yes	Yes	Yes
Portuguese	pt	Yes	Yes	Yes
Japanese	ja	Yes	Yes	Yes
Korean	ko	Yes	Yes	Yes
Mandarin Chinese	zh	Yes	Yes	Yes
Arabic	ar	Yes	Yes	Yes
Hindi	hi	Yes	Yes	Yes
Italian	it	Yes	Yes	Yes
Dutch	nl	Yes	Yes	Yes
Russian	ru	Yes	Yes	Yes
Turkish	tr	Yes	Yes	Yes
Polish	pl	Yes	Yes	Yes
Swedish	sv	Yes	Yes	Yes
Indonesian	id	Yes	Yes	Yes
Vietnamese	vi	Yes	Yes	Yes
Thai	th	Yes	Yes	Yes
Hebrew	he	Yes	Yes	Yes

Plus 20+ additional languages. Set "language": "auto" for automatic detection. Request "summary_language": "en" to receive AI highlights and summaries in English regardless of the spoken language.

Integration Guide

Get AI meeting notes running in your application with a few lines of code. The example below shows a complete Node.js integration that uploads a recording, waits for analysis, and extracts the action items. Python, Go, Ruby, and cURL examples are available in the full documentation.

meeting-notes.js — Node.js integration

import V100 from '@v100/sdk';

const v100 = new V100({ apiKey: process.env.V100_API_KEY });

// Analyze a meeting recording
const meeting = await v100.meetings.analyze({
  source: 'https://storage.example.com/standup.mp4',
  language: 'auto',
  features: {
    transcription: true,
    diarization: true,
    highlights: {
      action_items: true,
      decisions: true,
      questions: true,
      topics: true,
      sentiment: true
    },
    summary: true
  }
});

// Send action items to your project management tool
for (const item of meeting.highlights.action_items) {
  await createTask({
    title: item.text,
    assignee: item.assignee,
    dueDate: item.due,
    priority: item.priority,
    sourceUrl: `https://app.example.com/meetings/${meeting.meeting_id}?t=${item.timestamp}`
  });
}

// Post summary to Slack
await slack.chat.postMessage({
  channel: '#engineering',
  text: `*Meeting Summary*\n${meeting.summary}\n\n*Action Items:*\n${
    meeting.highlights.action_items
      .map(a => `- ${a.text} (${a.assignee}, due ${a.due})`)
      .join('\n')
  }`
});

1. Get an API Key

Sign up for a free V100 account. Your API key is available immediately in the dashboard. No credit card required for the free tier.

2. Send a Recording

POST the recording URL to /v1/meetings/analyze with the features you need. The API returns a meeting ID immediately while processing continues asynchronously.

3. Receive Results

Poll the meeting status endpoint or configure a webhook. When processing completes, the full transcript, AI highlights, summary, and export URLs are available in the response.

Tamper-Proof Transcripts with Dilithium Signatures

Every meeting transcript processed by V100 is cryptographically signed using ML-DSA-65 (FIPS 204), also known as Dilithium. This is a post-quantum digital signature algorithm that remains secure against both classical and quantum computers. The signature covers the complete transcript text, all word-level timestamps, speaker attributions, and AI-generated highlights. If a single character is altered after signing, verification fails.

Why Post-Quantum?

Meeting transcripts often serve as legal records, compliance artifacts, or contractual evidence. RSA and ECDSA signatures are vulnerable to future quantum computers running Shor's algorithm. ML-DSA-65 is a NIST-standardized lattice-based signature scheme that provides 128-bit security against quantum attackers. Transcripts signed today will remain verifiable for decades, even after large-scale quantum computers exist.

What Is Signed

The signature covers the SHA3-256 hash of the canonical JSON transcript, which includes every word with its timestamp and confidence score, all speaker labels, the AI-generated highlights, the summary, and the export file checksums. The public key and signature are included in the pq_signature field of the API response. You can verify the signature independently using any ML-DSA-65 implementation.

Verify transcript signature

curl -X POST https://api.v100.ai/v1/meetings/verify \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "meeting_id": "mtg_4d7e9a2f",
    "signature": "ml-dsa-65:0x3f8c1a..."
  }'

// Response
{
  "valid": true,
  "algorithm": "ML-DSA-65",
  "signed_at": "2026-03-28T14:22:18Z",
  "hash": "sha3-256:a4f8b2c1...",
  "covers": ["transcript", "highlights", "summary", "exports"]
}

Pricing

AI meeting notes are included in V100's platform pricing. Transcription and AI highlight extraction are bundled together -- there is no separate charge for the AI analysis layer on top of the transcript.

Free Tier

$0

10 hours/month of meeting analysis

All 40+ languages
Word-level timestamps
Speaker diarization
AI highlights + summary
SRT, VTT, JSON export

Growth

$0.006/min

Unlimited meeting analysis

Everything in Free
Real-time WebSocket streaming
Sentiment analysis
Transcript editing
Webhook callbacks

Enterprise

Custom

Volume discounts + SLA

Everything in Growth
PQ-signed transcripts (ML-DSA-65)
Custom highlight categories
HIPAA + BAA
Dedicated support

AI Meeting Notes API
Automated Summaries, Action Items & Transcription

What the AI Meeting Notes API Does

Real-Time vs Post-Meeting Analysis

Real-Time Mode

Post-Meeting Mode

AI-Detected Highlights

Action Items

Decisions

Questions

Topic Segmentation

Sentiment Analysis

Executive Summary

Supported Languages

Integration Guide

1. Get an API Key

2. Send a Recording

3. Receive Results

Tamper-Proof Transcripts with Dilithium Signatures

Why Post-Quantum?

What Is Signed

Pricing

Stop Taking Meeting Notes Manually

Explore Features

Video Transcription API

Post-Quantum Video Encryption

HIPAA-Compliant Video API

AI Meeting Notes APIAutomated Summaries, Action Items & Transcription

What the AI Meeting Notes API Does

Real-Time vs Post-Meeting Analysis

Real-Time Mode

Post-Meeting Mode

AI-Detected Highlights

Action Items

Decisions

Questions

Topic Segmentation

Sentiment Analysis

Executive Summary

Supported Languages

Integration Guide

1. Get an API Key

2. Send a Recording

3. Receive Results

Tamper-Proof Transcripts with Dilithium Signatures

Why Post-Quantum?

What Is Signed

Pricing

Stop Taking Meeting Notes Manually

Explore Features

Video Transcription API

Post-Quantum Video Encryption

HIPAA-Compliant Video API

AI Meeting Notes API
Automated Summaries, Action Items & Transcription