Diretta UPnP Renderer v2.1.4

The world's first native UPnP/DLNA renderer with Diretta protocol support - Low-Latency Edition

What's New in v2.1.4

Fix Audirvana local stream detection.

Link-local address detection — Audirvana streams via link-local addresses (169.254.x.x) were incorrectly detected as remote streams, enabling HTTP reconnection options that Audirvana's local HTTP server doesn't support. This caused playback interruptions, white noise, and track advancement issues with Audirvana Studio.
Resilient UPnP startup — On systems where the network isn't ready at boot (e.g., GentooPlayer with OpenRC), UPnP initialization now retries automatically instead of failing immediately.

Previous Versions

| Version | Highlights | |---------|-----------| | v2.1.3 | Fix target retry pre-check bypass | | v2.1.2 | Resilient target discovery (retry at startup instead of immediate exit) | | v2.1.1 | UAPP compatibility, format transition stability, high sample rate buffers, build capabilities logging | | v2.1.0 | Web Configuration UI, Advanced SDK settings, stop fix (herisson-88), libupnp auto-detect | | v2.0.6 | Advanced SDK settings, config migration, stop fix (herisson-88), libupnp auto-detect | | v2.0.5 | Stop fix for Holo Red (herisson-88), libupnp auto-detection, privilege drop removed | | v2.0.4 | Centralized logging, rebuffering on underrun, ARM NEON SIMD, systemd hardening, unit tests | | v2.0.3 | Audirvana compatibility (UPnP event deduplication), adaptive buffer for remote streaming | | v2.0.2 | Native DSD from Audirvana (DSDIFF parser), UPnP event notifications, gapless preload | | v2.0.1 | 24-bit white noise fix for DACs without 32-bit support | | v2.0.0 | Complete rewrite — low-latency architecture, lock-free ring buffer, AVX2 SIMD |

See CHANGELOG.md for full version history.

Support This Project

If you find this renderer valuable, you can support development:

Important notes:

Donations are optional and appreciated
Help cover test equipment and coffee
No guarantees for features, support, or timelines
The project remains free and open source for everyone

IMPORTANT - PERSONAL USE ONLY

This renderer uses the Diretta Host SDK, which is proprietary software by Yu Harada available for personal use only. Commercial use is strictly prohibited. See LICENSE for details.

Overview
Architecture
Features
Requirements
Upgrading
Quick Start
Supported Formats
Performance
Compatible Control Points
System Optimization
CPU Tuning
Command Line Options
Web Configuration UI
Troubleshooting
Documentation
Credits
License

Overview

This is a native UPnP/DLNA renderer that streams high-resolution audio using the Diretta protocol for bit-perfect playback. Unlike software-based solutions that go through the OS audio stack, this renderer sends audio directly to a Diretta Target endpoint (such as Memory Play, GentooPlayer, or hardware with Diretta support), which then connects to your DAC.

What is Diretta?

Diretta is a proprietary audio streaming protocol developed by Yu Harada that enables ultra-low latency, bit-perfect audio transmission over Ethernet. The protocol uses two components:

Diretta Host: Sends audio data (this renderer uses the Diretta Host SDK)
Diretta Target: Receives audio data and outputs to DAC (e.g., Memory Play, GentooPlayer, or DACs with native Diretta support)

Key Benefits

Bit-perfect streaming - Bypasses OS audio stack entirely
Ultra-low latency - ~300ms PCM buffer (vs ~1s in v1.x)
High-resolution support - Up to DSD1024 and PCM 1536kHz
Gapless playback - Seamless track transitions
UPnP/DLNA compatible - Works with any UPnP control point
Network optimization - Adaptive packet sizing with jumbo frame support

Architecture

Version 2.x uses a simplified, performance-focused architecture:

┌─────────────────────────────┐
│  UPnP Control Point         │  (JPlay, BubbleUPnP, mConnect, etc.)
└─────────────┬───────────────┘
              │ UPnP/DLNA Protocol (HTTP/SOAP/SSDP)
              ▼
┌───────────────────────────────────────────────────────────────┐
│  DirettaRendererUPnP v2.x                                     │
│  ┌─────────────────┐  ┌─────────────────┐  ┌───────────────┐  │
│  │   UPnPDevice    │─▶│ DirettaRenderer │─▶│  AudioEngine  │  │
│  │ (discovery,     │  │ (orchestrator,  │  │ (FFmpeg       │  │
│  │  transport)     │  │  threading)     │  │  decode)      │  │
│  └─────────────────┘  └────────┬────────┘  └───────┬───────┘  │
│                                │                   │          │
│                                ▼                   ▼          │
│                  ┌─────────────────────────────────────────┐  │
│                  │           DirettaSync                   │  │
│                  │  ┌───────────────────────────────────┐  │  │
│                  │  │       DirettaRingBuffer           │  │  │
│                  │  │  (lock-free SPSC, AVX2 convert)   │  │  │
│                  │  └───────────────────────────────────┘  │  │
│                  │              │                          │  │
│                  │              ▼ getNewStream() callback  │  │
│                  │  ┌───────────────────────────────────┐  │  │
│                  │  │      DIRETTA::Sync (SDK)          │  │  │
│                  │  └───────────────────────────────────┘  │  │
│                  └─────────────────────────────────────────┘  │
└───────────────────────────────────────────────────────────────┘
              │ Diretta Protocol (UDP/Ethernet)
              ▼
┌─────────────────────────────┐
│      Diretta TARGET         │  (Memory Play, GentooPlayer, etc.)
└─────────────┬───────────────┘
              ▼
┌─────────────────────────────┐
│            DAC              │
└─────────────────────────────┘

v2.0 vs v1.x Architecture

| Component | v1.x | v2.0 | |-----------|------|------| | SDK Base Class | DIRETTA::SyncBuffer | DIRETTA::Sync | | Data Model | Push (SDK manages timing) | Pull (getNewStream() callback) | | Ring Buffer | Standard | Lock-free SPSC with AVX2 | | Format Conversion | Per-sample | SIMD batch (8-32 samples) | | Thread Safety | Mutex-based | Lock-free atomics |

Features

Audio Quality

Bit-perfect streaming: No resampling or processing (when formats match)
PCM Bypass mode: Direct path for bit-perfect playback when no conversion needed
High-resolution support:
- PCM: Up to 32-bit/1536kHz
- DSD: DSD64, DSD128, DSD256, DSD512, DSD1024
Format support: FLAC, ALAC, WAV, AIFF, DSF, DFF, MP3, AAC, OGG
Gapless playback: Seamless album listening experience

Low-Latency Optimizations

Reduced buffers: 300ms PCM (was 1s), 800ms DSD
Micro-sleeps: 500µs flow control (was 10ms)
Lock-free path: Zero mutex in audio hot path
SIMD conversions: AVX2 for 8-32x throughput
Zero allocations: Pre-allocated buffers in steady state

UPnP/DLNA Features

Full transport control: Play, Stop, Pause, Resume, Seek
Device discovery: SSDP advertisement for automatic detection
Dynamic protocol info: Exposes all supported formats to control points
Position tracking: Real-time playback position updates

Network Optimization

Adaptive packet sizing: Synchronized with SDK cycle time
Jumbo frame support: Up to 16KB MTU for maximum performance
Automatic MTU detection: Configures optimal packet size

Robustness

Resilient startup: Retries target discovery indefinitely if target is not yet available, with periodic status logging
Auto-release: Diretta target released after idle timeout for coexistence with other Diretta hosts

Requirements

Supported Architectures

The renderer automatically detects and optimizes for your CPU:

| Architecture | Variants | Notes | |--------------|----------|-------| | x64 (Intel/AMD) | v2 (baseline), v3 (AVX2), v4 (AVX-512), zen4 | AVX2 recommended but not required | | ARM64 | Standard (4KB pages), k16 (16KB pages) | Pi 4/5 supported | | RISC-V | Experimental | riscv64 |

Note on older x64 CPUs: CPUs without AVX2 (Sandy Bridge, Ivy Bridge - 2011-2012) are fully supported. The build system automatically detects CPU capabilities and uses optimized scalar implementations when AVX2 is not available. Use the x64-linux-15v2 SDK variant for these systems.

Platform Support

| Platform | Status | |----------|--------| | Linux x64 | Supported (Fedora, Ubuntu, Arch, AudioLinux) | | Linux ARM64 | Supported (Raspberry Pi 4/5) | | Windows | Not supported | | macOS | Not supported |

Hardware

Minimum: Dual-core CPU, 1GB RAM, Gigabit Ethernet
Recommended: Quad-core CPU, 2GB RAM, 2.5/10G Et

DirettaRendererUPnP

Install / Use

README