Skip to content
jordanpeelePublic

About

AI production crew for live field journalism

Resources

Stars

1 star

Watchers

0 watching

Forks

Latest commit

 

History

2 Commits

Folders and files

Repository files navigation

Hiro Live

AI Production Crew for Live Field Journalism

Hiro Live is a multi-agent AI system that functions as a complete live broadcast production crew. It sits between a field camera and the audience, processing raw video through cloud OBS and adding AI-powered production elements: overlays, alerts, fact-checks, translations, and highlight clips.

The Vision

A single journalist with a camera, an AI production crew, and a wireless connection can produce live coverage that rivals a full broadcast team.

Agent Crew

Agent Role Responsibility
Orchestrator Executive Producer Coordinates all agents, owns state, resolves conflicts
Director Technical Director Controls OBS scenes, auto-BRB on feed drop
Graphics Engine Graphics Operator Renders overlays via WebSocket
Listener Transcriptionist STT, entity extraction, audio analysis
Fact-Checker Research Desk Real-time claim verification
Translator Interpreter Non-English speech translation
Clipper Highlights Producer Detect and extract highlights
Chat Wrangler Audience Producer Read chat, synthesize suggestions
Archivist Logger Record everything, generate reports

Quick Start

Prerequisites

  • Python 3.10+
  • Redis
  • OBS Studio (on cloud VM with GPU)
  • API keys: Anthropic (Claude), Deepgram

Installation

# Clone repository
git clone https://github.com/jordanpeele/hiro-live.git
cd hiro-live

# Create virtual environment
python -m venv venv
source venv/bin/activate

# Install dependencies
pip install -e ".[dev]"

# Copy environment template
cp .env.example .env
# Edit .env with your API keys

Running

# Start with mock OBS (for development)
python -m hiro_live.main --mock

# Start with real OBS connection
python -m hiro_live.main

Testing

# Run tests
pytest

# Run with coverage
pytest --cov=hiro_live

Build Stages

  1. MVP (2 weeks): Cloud OBS pipeline, chat alerts, auto-BRB
  2. Alpha (3-4 weeks): STT, lower thirds, chat wrangling, post-stream reports
  3. Beta (4-6 weeks): Fact-checking, translation, highlight clipping
  4. Production: Full nine-agent crew, continuous improvement

Architecture

Phone (Moblin) → SRT → Cloud OBS → RTMP → Twitch + YouTube
                         ↓
                   Hiro Live Agents
                         ↓
              WebSocket → OBS Browser Sources (Overlays)

Documentation

Project Status

Current Stage: MVP (Stage 1)
Target: First live test stream with Ben Jenkins

License

MIT


Coby Worldwide | March 2026

About

AI production crew for live field journalism

Resources

Stars

1 star

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages