Multi-Camera Livestream Framework

Professional multi-camera 4K live streaming pipeline with Dante audio synchronization. Open documentation. Any budget.

$3K to $50K+

macOS + Apple Silicon

Dante Audio Sync

4 Tested Profiles

Quick Start

Profiles

Why This Exists

The Problem This Framework Solves
Technical Architecture
Choose Your Profile
Installation and Quick Start
Configuration System
Operational Workflow
Features and Capabilities
Key Statistics
Real-World Deployments
Documentation Map
Cross-Organ Context
Roadmap
Community and Research
Contributing
License
Author

The Problem This Framework Solves

Every multi-camera streaming guide ends the same way: “…and then figure out audio sync yourself.”

You have spent thousands on cameras, capture hardware, and a capable Mac. OBS is installed. Dante Controller is open. The hardware is physically connected. Now what?

How do you route Dante audio into OBS without drift?
Why is audio three frames behind video after twenty minutes?
What encoder settings prevent the stream from dying mid-event?
How does a volunteer operator reproduce your exact setup next week?
Where is the single document that ties cameras, audio network, encoding, and streaming into one coherent pipeline?

That document does not exist. Until now.

What Is Missing from Existing Solutions

Problem	Existing Resources	This Framework
Hardware selection for a specific budget	Scattered forum posts	4 tested profiles ($3K to $50K+) with full BOMs
Dante audio routed into OBS	Fragmented guides across 3 vendors	Complete clock-locked architecture with signal flow diagrams
Audio-video synchronization over long broadcasts	“Good luck”	Global sample clock via Dante, measured drift <50ms over 1 hour
Volunteer operator handoff	“Watch this 45-minute video”	8-phase runbook with decision-tree troubleshooting
Reproducible streaming setup for academic citation	Nothing	BibTeX-ready, version-locked, config-as-code profiles
Pre-stream system validation	Manual checklists	Automated health-check scripts with JSON output

How This Framework Differs from Products

Capability	ATEM Mini	vMix	OBS Alone	This Framework
Audio synchronization	HDMI embedded only	Windows audio stack	Manual configuration	Dante global clock
Budget flexibility	Fixed hardware tier	Fixed software license	Unknown starting point	4 documented profiles
Operational documentation	Product manual	Tutorial videos	Community wiki fragments	95% operational coverage
Volunteer-ready operation	“Watch the video”	“Watch the video”	“Figure it out”	8-phase runbook
Reproducible configuration	“Buy same hardware”	“Same version”	No mechanism	Config-as-code YAML profiles
Pre-stream health checks	None	None	None	Automated shell scripts

ATEM and vMix are products. OBS is software. This framework is the knowledge layer that makes the entire pipeline work together. It documents not just what to connect, but how, why, and what to do when something breaks.

Technical Architecture

The pipeline separates concerns into five independent subsystems connected via standardized protocols. Each subsystem can be understood, debugged, and upgraded independently.

Signal Flow

┌──────────────────────────────────────────────────────────────┐
│  INPUT LAYER                                                 │
│                                                              │
│  Cameras (HDMI out)           Remote Callers (NDI)           │
│  ├─ Camera 1 ──┐              └─ OBS NDI stream ──┐         │
│  ├─ Camera 2 ──┤                                   │         │
│  ├─ Camera 3 ──┤  DeckLink Quad HDMI               │         │
│  └─ Camera 4 ──┘  (Echo Express SE I / TB3)        │         │
│        │                                            │         │
├────────┼────────────────────────────────────────────┼─────────┤
│  CAPTURE + SYNC LAYER                              │         │
│        │                                            │         │
│        ▼                                            ▼         │
│  OBS Studio ◄──────── Audio ──── MOTU 8PRE-ES      │         │
│  (4× DeckLink inputs)          (Dante I/O)         │         │
│  (NDI caller inputs)              │                 │         │
│                                   │                 │         │
│  Dante Network ◄─────────────────┘                  │         │
│  ├─ 48 kHz global sample clock                      │         │
│  ├─ AVIO nodes (camera audio converters)            │         │
│  └─ Dedicated managed Ethernet switch               │         │
│                                                              │
├──────────────────────────────────────────────────────────────┤
│  PROCESSING LAYER                                            │
│                                                              │
│  Ableton Live (Dante clock master)                           │
│  ├─ Input: Camera feeds via Dante                            │
│  ├─ Mixing: Per-channel EQ, gain, effects                    │
│  ├─ Clock: Drives all Dante devices at 48 kHz                │
│  └─ Output: Stereo master → MOTU → OBS                      │
│                                                              │
│  Blender (optional graphics)                                 │
│  ├─ Input: Generative parameters via MIDI                    │
│  └─ Output: Screen share → OBS                               │
│                                                              │
├──────────────────────────────────────────────────────────────┤
│  OUTPUT LAYER                                                │
│                                                              │
│  OBS Encoder (H.264 / H.265)                                │
│  ├─ Video: 8–25 Mbps (codec-dependent)                      │
│  ├─ Resolution: 4K or 1080p mosaic                           │
│  ├─ Framerate: 30fps or 60fps                                │
│  └─ Protocol: RTMP                                           │
│        │                                                     │
│        ▼                                                     │
│  YouTube Live / Twitch / Custom RTMP                         │
│  Local recording (continuous backup)                         │
│                                                              │
└──────────────────────────────────────────────────────────────┘

Clock Synchronization — The Core Innovation

Audio-video drift is the central unsolved problem in multi-camera streaming. The framework eliminates it through a global clock hierarchy:

Ableton Live (Master Clock, 48 kHz)
    │
    ▼  sync
Dante Network (all devices phase-locked, <1µs deviation)
    │
    ▼  sync
Camera audio converters (AVIO nodes lock to Dante clock)
    │
    ▼  result
Zero cumulative drift across all audio sources

Every audio device in the pipeline — microphones, camera audio feeds, remote caller returns — is phase-locked to a single 48 kHz sample clock distributed over the Dante audio network. The MOTU 8PRE-ES serves as the primary Dante I/O interface, with Ableton Live acting as the clock master. This architecture means that audio from Camera 1, Camera 4, and a remote NDI caller all arrive at OBS with identical timing. No post-production correction required.

Scaling Architecture

The framework scales from a single Mac to a multi-room production facility:

Phase	Setup	Cameras	Key Addition
Phase 1 (current)	Single M1 Ultra, all local	4 tested, 8 expandable	DeckLink Quad HDMI
Phase 2	Multi-room, NDI feeds to master	8–10	M1 Mac mini per room, NDI aggregation
Phase 3	Building-scale	10–15 rooms	Redundant Dante switches, NAS archive, analytics

Choose Your Profile

The framework ships with four tested hardware profiles spanning a 15x cost range. Each profile includes a complete bill of materials, configuration files, and operational documentation.

┌─────────────────┐  ┌─────────────────┐  ┌─────────────────┐  ┌─────────────────┐
│    BUDGET       │  │    MOBILE       │  │    STUDIO       │  │   BROADCAST     │
│     ~$3K        │  │     ~$8K        │  │    ~$20K        │  │    $50K+        │
├─────────────────┤  ├─────────────────┤  ├─────────────────┤  ├─────────────────┤
│ 2 cameras       │  │ 2–4 cameras     │  │ 4 cameras       │  │ 8+ cameras      │
│ USB capture     │  │ TB3 capture     │  │ DeckLink PCIe   │  │ Multi-DeckLink  │
│ USB audio       │  │ Portable Dante  │  │ Full Dante      │  │ Dante + SDI     │
│ Mac mini        │  │ MacBook Pro     │  │ Mac Studio      │  │ Mac Studio/Pro  │
├─────────────────┤  ├─────────────────┤  ├─────────────────┤  ├─────────────────┤
│ Best for:       │  │ Best for:       │  │ Best for:       │  │ Best for:       │
│ • Worship       │  │ • Touring       │  │ • Research      │  │ • Esports       │
│ • Classroom     │  │ • Remote prod   │  │ • Corporate     │  │ • Enterprise    │
│ • Simple events │  │ • Venue-hopping │  │ • Full-featured │  │ • Multi-room    │
└─────────────────┘  └─────────────────┘  └─────────────────┘  └─────────────────┘

Profiles are YAML configuration files that override a shared defaults layer. The configuration system computes minimum hardware requirements from production needs (number of cameras, resolution, audio channels) using scaling formulas, then validates that your hardware meets those requirements. See Configuration System for details.

Quick Start

Prerequisites

macOS 13.0 or later (tested on 14.2.1)
Apple Silicon Mac (M1/M2/M3 — Studio, mini, or MacBook Pro)
Gigabit Ethernet connection
Homebrew package manager

Installation

# Clone the repository
git clone https://github.com/organvm-iii-ergon/multi-camera--livestream--framework.git
cd multi-camera--livestream--framework

# Install build dependencies (yq for YAML parsing, gettext for templating)
make install-deps

# Generate configuration for your hardware profile
make config PROFILE=studio    # Options: budget, mobile, studio, broadcast

# Verify your system meets requirements
./software/scripts/setup-macos.sh

# Run pre-stream health check
./software/scripts/health-check.sh

First Broadcast — Five-Day Onboarding

Day	Task	Reference Document
1–2	Verify hardware against profile BOM	hardware/COMPATIBILITY.md
3–4	Configure Dante audio network and clock sync	docs/AUDIO-DANTE.md
5	Full dry run using the 8-phase runbook	docs/RUNBOOK.md
6–7	Go live with local recording (no public stream)	docs/STREAMING.md

Makefile Commands

make config                 # Generate config from default profile (studio)
make config PROFILE=mobile  # Generate config from a specific profile
make docs                   # Generate documentation from templates
make all                    # Generate both config and docs
make switch-profile PROFILE=budget  # Switch active profile
make list-profiles          # List available profiles
make validate               # Syntax-check all shell scripts
make health-check           # Run pre-stream health check
make clean                  # Remove all generated files

Configuration System

The framework uses a production-centric configuration architecture where YAML profiles describe what your show needs, and the system computes whether your hardware can deliver it.

Architecture

software/configs/
├── defaults.yaml              # Base config inherited by all profiles
├── profiles/
│   ├── studio.yaml           # M1 Mac Studio + DeckLink + full Dante
│   ├── mobile.yaml           # MacBook Pro + portable gear
│   ├── budget.yaml           # Mac mini + USB capture
│   └── broadcast.yaml        # Mac Studio/Pro + multi-DeckLink + SDI
└── active -> profiles/studio.yaml   # Symlink to active profile

software/generated/            # Output (gitignored)
├── config.sh                 # Shell-sourceable variables
├── config.json               # JSON for external tools
└── docs/                     # Generated documentation

Profile Ontology

Each profile is organized into three layers:

Production requirements — what the show needs (cameras, resolution, audio channels, stream destinations)
Hardware specification — what your equipment provides (detected by category, not brand name)
Software configuration — applications identified by capability, not product name

This means the framework identifies hardware by role (e.g., pcie_capture, dante_interface, streaming_switcher) rather than brand. Detection uses pattern matching against system device lists, making the framework extensible to hardware from any vendor.

Scaling Formulas

The configuration system computes minimum requirements from production needs:

# RAM: base + (cameras * per_camera)
# Example: 4 cameras = 8 + (4 * 4) = 24 GB minimum
scaling:
  compute:
    ram_base_gb: 8
    ram_per_camera_gb: 4
    ram_per_4k_camera_gb: 8

  network:
    bandwidth_base_mbps: 100
    bandwidth_per_camera_mbps: 150    # ~150 Mbps per 4K30 stream

Using Configuration in Scripts

All shell scripts load configuration with graceful fallbacks:

SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
if [[ -f "$SCRIPT_DIR/../lib/config-utils.sh" ]]; then
    source "$SCRIPT_DIR/../lib/config-utils.sh"
    load_config 2>/dev/null || true
fi

# Use config values with sensible defaults
MIN_DISK="${CONFIG_THRESHOLDS_MIN_DISK_SPACE_GB:-50}"
AUDIO_MODEL="${CONFIG_HARDWARE_AUDIO_INTERFACE_MODEL:-8PRE-ES}"

Creating a Custom Profile

# Copy the closest existing profile
cp software/configs/profiles/studio.yaml software/configs/profiles/mysetup.yaml

# Edit with your hardware and production requirements
$EDITOR software/configs/profiles/mysetup.yaml

# Generate config and validate
make config PROFILE=mysetup
./software/scripts/health-check.sh

Operational Workflow

Shell Scripts

The framework includes four operational scripts that automate the entire broadcast lifecycle:

Script	Purpose	When to Run
`setup-macos.sh`	System verification — checks macOS version, drivers, hardware detection	Once after installation
`health-check.sh`	Pre-stream diagnostics — validates audio devices, capture hardware, disk, CPU, network	30–45 minutes before every broadcast
`launch-studio.sh`	Application startup — launches all apps in correct dependency order	Start of broadcast session
`shutdown-studio.sh`	Graceful shutdown — quits apps via AppleScript, hardware shutdown reminders	End of broadcast session

The 8-Phase Runbook

The operational runbook structures every broadcast into eight sequential phases. Each phase includes a checklist, timing estimate, and troubleshooting decision tree for common failures.

Phase	Duration	Focus
1	10 min	Network and Dante — Ethernet, device discovery, clock sync
2	10 min	Audio system — MOTU power, Dante routing, Ableton project load
3	10 min	Video capture — DeckLink verification, camera connections, OBS sources
4	5 min	Scene composition — OBS scenes, transitions, overlays
5	5 min	Streaming configuration — RTMP keys, bitrate, recording path
6	5 min	Remote callers — NDI input verification, audio return
7	5 min	Final checks — health-check script, test recording, audio levels
8	—	Go live — start stream, monitor, graceful shutdown

The runbook is written for volunteer operators with no prior streaming experience. A trained volunteer can complete Phases 1–7 in under 45 minutes.

Health Check Output

The health-check script supports both human-readable and machine-parseable output:

# Human-readable (default)
./software/scripts/health-check.sh

# Verbose mode with detailed information per check
./software/scripts/health-check.sh --verbose

# JSON output for automation and monitoring
./software/scripts/health-check.sh --json

Exit codes: 0 = all checks passed, 1 = critical failures, 2 = warnings (can proceed).

Features and Capabilities

Hardware Stack (Studio Profile)

Component	Role	Specification
M1 Mac Studio (Ultra)	Compute, encoding, streaming	128 GB RAM, 20 cores, NVMe storage
DeckLink Quad HDMI	Video capture (4 inputs)	4K30 per input, <1ms hardware capture latency
Echo Express SE I	Thunderbolt 3 PCIe chassis	Houses DeckLink card
MOTU 8PRE-ES	Dante audio I/O	24 in / 28 out, 48 kHz–96 kHz sample rates
Dante AVIO USB	Camera audio bridge	Converts USB audio devices to Dante network
Managed Ethernet switch	Dante audio backbone	Dedicated VLAN, gigabit, low jitter

Software Stack (Version-Locked)

Application	Role	Pinned Version
OBS Studio	Video mixing, encoding, RTMP streaming	29.1.3
Ableton Live	Audio mixing, Dante clock master	12.0.10
Dante Controller	Audio network routing and monitoring	4.5.3
Blender (optional)	Real-time 3D graphics overlay	—
DeckLink Driver	Video capture device driver	13.2 (arm64)
MOTU Driver	Audio interface driver	2.23.5

Version pinning ensures reproducibility. The exact combination of macOS, drivers, and applications listed above has been tested for stability under sustained 4K streaming loads.

NDI Caller Integration

Remote participants connect via NDI (Network Device Interface) with <200ms latency on local networks and <500ms over the internet. Each caller runs a local OBS instance streaming via NDI to the master machine. Audio return is routed through the Dante network for clock-synchronized monitoring.

Documentation Coverage

The repository contains 35+ files and over 25,000 words of professional technical documentation:

Document	Content
ARCHITECTURE.md	System design, signal flow, scaling roadmap
RUNBOOK.md	8-phase operational checklist
AUDIO-DANTE.md	Dante network setup, clock hierarchy, troubleshooting
VIDEO-CAPTURE.md	DeckLink configuration, HDMI specifications
NDI-CALLERS.md	Remote caller onboarding, network requirements
SOFTWARE.md	Version-pinned installation steps
STREAMING.md	RTMP platform configuration (YouTube, Twitch)
TROUBLESHOOTING.md	Decision-tree diagnostics
FAQ.md	Common questions and gotchas
BENCHMARKS.md	Performance measurements
MONITORING.md	Runtime observability

Key Statistics

Metric	Value
Maximum simultaneous cameras	4 tested, 8+ expandable
Capture-to-stream latency	2–3 seconds (platform-dependent)
Caller return latency	<200ms (Dante local), <500ms (WebRTC)
Audio sync drift	<50ms measured over 1 hour
Sustained CPU load	50–60% (M1 Ultra, 4K, full pipeline)
Sustained GPU load	65–70% (H.265 encoding)
Network jitter (Dante)	<2ms on dedicated switch
Dropped frames (2-hour test)	0
Documentation coverage	~95% of operational procedures

Real-World Deployments

House of Worship

Profile: Budget ($3K)

Challenge: Volunteer operators rotate weekly

“Zero audio sync complaints after implementing Dante. Volunteer onboarding went from 3 training sessions to 1 because of the runbook.”

University Research Lab

Profile: Studio ($20K)

Challenge: Reproducibility for academic publications

“Our streaming setup has been cited in 3 papers. Version-locked configs mean we can reproduce results from 2 years ago.”

Touring Performance Artist

Profile: Mobile ($8K)

Challenge: Different venue every night

“Setup time dropped from 4 hours to 45 minutes. I have venue profiles for 15 theaters now.”

Corporate Events

Profile: Studio ($20K)

Challenge: Replaced $15K-per-event vendor

“Third event paid for the entire setup. NDI callers look as good as local cameras.”

See all deployment scenarios in docs/USE-CASES.md.

Documentation Map

"I want to..."

├── BUILD A SYSTEM
│   ├── Understand the architecture ──→ docs/ARCHITECTURE.md
│   ├── Buy hardware ─────────────────→ hardware/BOM.csv
│   ├── Check compatibility ──────────→ hardware/COMPATIBILITY.md
│   └── Install software ────────────→ docs/SOFTWARE.md
│
├── CONFIGURE AUDIO
│   ├── Set up Dante ─────────────────→ docs/AUDIO-DANTE.md
│   ├── Configure cameras ────────────→ hardware/CAMERA-SETTINGS.md
│   └── Add remote callers ──────────→ docs/NDI-CALLERS.md
│
├── GO LIVE
│   ├── Pre-stream checklist ─────────→ docs/RUNBOOK.md
│   ├── Configure streaming ──────────→ docs/STREAMING.md
│   └── Run health check ────────────→ software/scripts/health-check.sh
│
├── FIX SOMETHING
│   ├── Troubleshooting ──────────────→ docs/TROUBLESHOOTING.md
│   └── FAQ ──────────────────────────→ docs/FAQ.md
│
├── AUTOMATE
│   ├── Generate configuration ───────→ make config PROFILE=<name>
│   ├── Launch all applications ──────→ software/scripts/launch-studio.sh
│   └── Shut down gracefully ─────────→ software/scripts/shutdown-studio.sh
│
└── UNDERSTAND THE PROJECT
    ├── Comparison to alternatives ───→ docs/COMPARISON.md
    ├── Real-world deployments ───────→ docs/USE-CASES.md
    ├── Performance benchmarks ───────→ docs/BENCHMARKS.md
    └── Roadmap ──────────────────────→ ROADMAP.md

Cross-Organ Context

This repository lives within ORGAN-III (Ergon) — the commerce and product organ of the organvm system. ORGAN-III houses tools, products, and frameworks intended for real-world deployment and practical use.

The multi-camera livestream framework connects to other organs in the system:

ORGAN-I (Theoria) — The theoretical foundations that inform this framework’s design. The production-centric ontology in the configuration system (hardware identified by role and capability rather than brand) reflects ORGAN-I’s work on category-theoretic abstraction and ontological design. The scaling formulas that derive minimum requirements from production needs are an applied instance of ORGAN-I’s recursive specification patterns.
ORGAN-II (Poiesis) — The art and performance organ. This framework was originally built to support live-streamed artistic performances — generative music broadcasts, multi-camera documentation of experimental theatre, and real-time visual art events. The Blender graphics pipeline (MIDI-driven generative overlays from Ableton into OBS via screen share) is a direct collaboration point between ORGAN-II’s creative output and ORGAN-III’s production infrastructure. The metasystem-master project in ORGAN-II represents the artistic vision that this framework makes technically possible.
ORGAN-IV (Taxis) — The orchestration organ. The framework’s profile-based configuration system and automated health checks exemplify the governance patterns coordinated by ORGAN-IV. The CI/CD workflows specified in the GitHub Actions configuration follow ORGAN-IV’s cross-repository validation standards.
ORGAN-V (Logos) — The public process organ. The framework’s emphasis on open documentation, reproducibility, and academic citation readiness aligns with ORGAN-V’s mission of building in public. The research directory (research/METHODOLOGY.md, research/PUBLICATION.md, references.bib) is structured for ORGAN-V essay integration.

Roadmap

Current: v1.0.0

Core documentation complete (Architecture, Runbook, Troubleshooting, FAQ)
Profile-based configuration system (YAML profiles, shell/JSON generation)
Shell scripts (setup, health-check, launch, shutdown)
Studio profile verified on M1 Mac Studio (Ultra)
GitHub templates (issues, PRs, hardware compatibility reports)

Near-Term

Timeline	Focus
Next	Budget and Mobile profile validation, video walkthroughs
Following	8-camera support (dual DeckLink), OBS 30.x compatibility, macOS 15 testing
After that	Automation (auto-switching via audio levels), health monitoring daemon, Slack/Discord alerts

Long-Term Vision

Goal	Description
Multi-studio federation	Connect 10–15 rooms into a single production
AI camera selection	Automatic switching hints based on audio levels and motion
SMPTE 2110 support	Broadcast-standard IP video
Linux profile	Ubuntu-based deployments for enterprise

See the full roadmap with quarterly breakdowns in ROADMAP.md.

Community and Research

Academic Citation

This framework is designed for reproducibility in academic and professional contexts. All configurations are version-locked and citable:

@misc{mcls-framework-2025,
  author       = {4444j99},
  title        = {Multi-Camera Livestream Framework},
  year         = {2025},
  url          = {https://github.com/organvm-iii-ergon/multi-camera--livestream--framework},
  note         = {Open-source documentation for reproducible multi-camera streaming setups}
}

Research Materials

The research/ directory contains:

METHODOLOGY.md — Research methodology and reproducibility framework
PUBLICATION.md — Publication strategy and target venues
references.bib — BibTeX bibliography of related work

Contributing

This is a living documentation project. Contributions across several areas are welcome:

Hardware testing — Test your setup and submit a hardware compatibility report using the issue template
Documentation improvements — Clarify unclear sections, fix errors, improve diagrams
New profiles — Add configurations for untested hardware combinations
Alternative hardware — Document non-Blackmagic capture cards, non-MOTU Dante interfaces
Video content — Record walkthrough tutorials, troubleshooting demonstrations
Translation — Localize core documentation for non-English-speaking communities

See the pull request template and issue templates for contribution guidelines.

License

Documentation: CC-BY-4.0 (Creative Commons Attribution)
Code and Scripts: MIT

Author

@4444j99 — Part of the organvm system.

Built for multimedia artists, academic researchers, live performance practitioners, and anyone who needs professional multi-camera streaming without a professional budget.

Professional streaming. Open documentation. Your budget.

Architecture · Runbook · Audio/Dante · Troubleshooting · FAQ · Roadmap

Multi-Camera Livestream Framework

Table of Contents

The Problem This Framework Solves

What Is Missing from Existing Solutions

How This Framework Differs from Products

Technical Architecture

Signal Flow

Clock Synchronization — The Core Innovation

Scaling Architecture

Choose Your Profile

Quick Start

Prerequisites

Installation

First Broadcast — Five-Day Onboarding

Makefile Commands

Configuration System

Architecture

Profile Ontology

Scaling Formulas

Using Configuration in Scripts

Creating a Custom Profile

Operational Workflow

Shell Scripts

The 8-Phase Runbook

Health Check Output

Features and Capabilities

Hardware Stack (Studio Profile)

Software Stack (Version-Locked)

NDI Caller Integration

Documentation Coverage

Key Statistics

Real-World Deployments

House of Worship

University Research Lab

Touring Performance Artist

Corporate Events

Documentation Map

Cross-Organ Context

Roadmap

Current: v1.0.0

Near-Term

Long-Term Vision

Community and Research

Academic Citation

Research Materials

Contributing

License

Author