Roadmap
What's being actively built, what's coming next, and what has already shipped. Timelines are estimates and may shift as development progresses.
Platform Compatibility
Here's a summary of what's currently supported as of REFEREE version latest. Features not currently supported are currently in development or planned for a future release as this is the highest priority.
Windows
Primary desktop target with the most complete REFEREE feature support.
| Platform OS / Environment | GPU GPU hardware vendor | AI Upscaling Resolution enhancement | Motion Boost Frame rate interpolation | HDR Mapping SDR → HDR tone mapping | REFEREE Relay Relay to a peer device |
|---|---|---|---|---|---|
| Windows .exeWindows 10 / 11 | |||||
LinuxEarly development
Linux support is still early
This platform may run into more issues than its Windows counterpart. These builds are currently best suited for REFEREE Relay workflows where this device connects to, or is paired with, a Windows-based REFEREE peer for the most complete processing support.
| Platform OS / Environment | GPU GPU hardware vendor | AI Upscaling Resolution enhancement | Motion Boost Frame rate interpolation | HDR Mapping SDR → HDR tone mapping | REFEREE Relay Relay to a peer device |
|---|---|---|---|---|---|
| Linux .debUbuntu 22.04+ · Debian 12+ .rpmFedora 38+ · RHEL / CentOS Stream 9+ | |||||
DockerEarly development
Docker support is still early
This platform may run into more issues than its Windows counterpart. These builds are currently best suited for REFEREE Relay workflows where this device connects to, or is paired with, a Windows-based REFEREE peer for the most complete processing support.
| Platform OS / Environment | GPU GPU hardware vendor | AI Upscaling Resolution enhancement | Motion Boost Frame rate interpolation | HDR Mapping SDR → HDR tone mapping | REFEREE Relay Relay to a peer device |
|---|---|---|---|---|---|
| Docker nvidiaNVIDIA Container Toolkit required amdAMD Mesa or ROCm required | |||||
Supported Coming soon Not supportedAI AI-accelerated
In Progress
Features currently under active development.
AMD AMF Support on Windows
In ProgressFull hardware-accelerated upscaling and frame generation on AMD RX 7000 and RX 8000 series GPUs via AMD AMF integration.
Docker Container Support
In ProgressExpanded Docker profile support to allow running REFEREE in a containerized environment, with separate NVIDIA and AMD configurations.
Planned for the Future
Features on the roadmap that are not yet in active development.
macOS Support
PlannedNative macOS application with Metal-accelerated upscaling and Apple Silicon support, bringing REFEREE to Mac.
Intel Arc GPU Support
PlannedHardware-accelerated upscaling and processing on Intel Arc discrete GPUs via the XeSS and oneVPL APIs.
Have a feature request?
Open an issue or start a discussion on GitHub. Community feedback directly shapes the priority of items on this roadmap.
Open an issue on GitHubChangelog
A summary of what shipped in each release.
v0.6.0
April 27, 2026- +Server module split — auth, relay, routing, status, stream handling, state, and response helpers are now organized into focused modules with expanded coverage
- +API token initialization now uses the resolved app data directory more consistently for desktop and headless server startup
- +Relay readiness checks now verify the linked peer before routing streams and report clearer offline or not-ready states
- +Remote Relay settings lock automatically while the linked peer is offline, with inline status and error messages in the settings UI
- +Motion Boost can now be toggled when controlling a NVIDIA-backed relay peer
- +Status and settings refresh paths were tuned to reduce redundant UI work while keeping relay state current
- +Developer docs now cover live stream settings updates, HLS proxy segment extension guidance, and improved local demo setup notes
- +Project licensing clarified as GPL-3.0-or-later with updated README guidance
v0.5.0
April 25, 2026- +Relay system — link REFEREE instances on the local network and route streams through a linked peer with persistent credential storage
- +Remote-backed sessions — the local player proxies HLS output from the linked relay peer with seamless handoff
- +Real-time relay link status — connection availability and peer reachability are monitored and surfaced live in the UI
- +Relay stream control — stream start/stop commands and heartbeat are forwarded to the relay peer; stream settings are locked when controlled remotely
- +Auto-relay mode — relay routing is automatically activated when the local GPU does not meet processing requirements
- +Remote stream settings — stream settings can be configured on the relay peer directly from the controlling instance
- +Relay UI — trusted peer panel, per-peer status badges, and inline relay status indicators in the settings card
- +
/v1/pingAPI endpoint for liveness checks and LAN peer discovery - +Audio timing stability improvements merged across all playback paths
- +Stream settings update refactored for cleaner error reporting
v0.4.3
April 24, 2026- +Native C++/Rust RIFE worker — frame interpolation rewritten as a native worker, replacing the Python-based implementation for lower latency and no runtime dependency
- +FidelityFX CAS sharpening — Contrast Adaptive Sharpening integrated across all executor paths with a custom GLSL shader variant
- +AMD AMF hardware-accelerated decoding — universal executor now uses AMF for hardware decode on AMD GPUs
- +Hardware-accelerated denoising — hqdn3d denoising added to native and universal source-pull paths; switched to GPL FFmpeg builds to enable filter
- +NVIDIA FRUC improvements — constant frame rate enforcement, dynamic target FPS, and corrected transcode pacing
- +Built-in HLS stream player — embedded player window launches automatically when a stream starts
- +@cmacrowther/referee-client package — new npm package providing a typed API client and React integration hooks
- +Rate limiting and SSRF protection added to the server
- +LAN peer discovery — initial foundation for discovering REFEREE instances on the local network
- +HLS manifest rewriting now selects the highest bandwidth video variant
- +Fixed audio/video sync and stabilized NVIDIA encoding path under live HLS conditions
- +Fixed Docker volume ownership for correct container runtime permissions
v0.4.1
April 19, 2026- +AMD executor — full AMD pipeline on Windows with VCEEncC command building, hardware upscale support, and capability detection
- +RIFE frame interpolation — streaming RIFE support via embedded Python runtime and dedicated RIFE wrapper
- +Portable interpolation fallback — unsupported interpolation cases are routed to a software fallback with gap handling
- +Non-HLS source support — pipeline staged launch and preprocess steps now handle non-HLS inputs
- +Binary manager overhaul — unified binary resolution for RIFE, FFmpeg, and shader management
- +Intermediate execution plan — new intermediate plan structure for cleaner pipeline orchestration
- +libplacebo preprocess filter plan — dedicated preprocess stage for libplacebo-accelerated processing
- +HDR preprocess ownership flag — prevents double-processing of HDR operations in the Universal executor
- +Engine settings toggle in debug panel — new section with per-engine controls and enhanced debug UI
- +AI badges in settings — upscaling feature cards now show AI and hardware acceleration indicators
- +Fixed VCEEncC AMF encoder retry logic on Linux
- +Fixed RIFE Vulkan library detection in FFmpeg preprocessor
v0.3.1
April 18, 2026- +Source content classification — automatic detection of animated vs. live-action content to select the optimal upscaler
- +ArtCNN shader support — ArtCNN C4F16 GLSL shader added as an additional upscaling option
- +Universal shader manager — shader installation refactored to support multiple shader types beyond Anime4K
- +Stream info badges now display active content kind and selected upscaler
v0.3.0
April 12, 2026- +Pipeline rearchitecture — staged execution model with discrete preprocess, normalize, encode, and HLS package stages for improved reliability and extensibility
- +Universal executor — new cross-platform backend replaces the previous platform-specific executors, with a dedicated NVIDIA-specialized executor for Windows-exclusive features
- +Anime4K GLSL shader upscaling — 2x upscale path via Anime4K shaders for anime content
- +NVIDIA AI processing option — explicit GPU utilization preference exposed in settings
- +FFmpeg binary management — bundled FFmpeg support with ffprobe integration for improved cross-platform compatibility
- +HLS source relay — new source module proxies HLS inputs directly into the pipeline without re-muxing
- +HLS packager — dedicated FfmpegHlsPackager stage; segment list size increased to 8; HEVC codec tag added for Apple device compatibility
- +Source/upscaled video playback synchronisation — source and upscaled streams are kept in sync during live viewing
- +Settings UI redesigned — tabbed navigation layout with CapabilityChip indicators and a ProcessingPathSummary for at-a-glance pipeline status
- +Improved stderr handling — StderrMode propagated through pipeline stages for structured error capture and logging
- +Pipeline session cleanup — completion signalling ensures sessions are fully torn down before a new pipeline starts
- +Fixed Vulkan ICD library path in NVIDIA Docker image
v0.2.1
April 11, 2026- +Improved playlist session management — existing sessions are now cleanly stopped before a new stream is started
- +Fixed Cargo default-run entry to correctly target the server binary
v0.2.0
April 5, 2026- +Docker support — separate NVIDIA and AMD profiles with dedicated Dockerfiles
- +Headless server mode — run REFEREE without a desktop UI for self-hosted and container deployments
- +NVIDIA GPU detection for headless mode; encoder binary handling refactored
v0.1.8
April 4, 2026- +Fixed Windows and Linux artifact targets in the release workflow
- +Tauri build targets now explicitly specified for correct package output
v0.1.7
April 4, 2026- +External player detection and launch — configure and open streams directly in a local media player from the UI
- +Stream source name now displayed in the stream info badge
- +Lookahead argument added for both NVENC and VCE encoder backends
- +Removed Windows-specific pipeline configuration that could cause cross-platform issues
- +Frame generation and HDR toggles fixed; stream card now shows Open button once a stream is active
- +License updated to GNU GPL v3