Changelog & Release Notes

All notable changes, bug fixes, and feature additions to the now2sdk framework will be documented on this page.

Version 0.1.1 (July 24, 2026)

New Features & APIs

• LFile Property-Based Daily Event & Crash Logging: Added thread-safe 24-hour daily event and crash logging controlled via setProps (log, log.path, log.name, log.file). Supports automatic home directory expansion (~), fallback to ./logs/, and logs playout commands (PLAY, PAUSE, STOP, NEXT, BACK) alongside fatal OS crash signals.
• LRecorder Auto-Split & Manual Split: Added time-based automatic file splitting (setSplitMinute(int minutes)) and instant manual file splitting (Split()) to LRecorder. Recording files can automatically rotate every N minutes or on demand without losing audio/video frames or stopping the recording pipeline. Filenames are automatically indexed (filename_001.mp4, filename_002.mp4). Also exported via C API (now2sdk_recorder_set_split_minute, now2sdk_recorder_split).
• Full Medialooks Broadcast Matrix & Profile Matching: Updated LRecorder with a complete broadcast and web container/codec matrix (mp4, mov, mxf, mxf_d10, avi, mkv, webm, ts, mpg, dv, gxf). Integrated Avid DNxHR profile aliases (dnxhr_lb, dnxhr_sq, dnxhr_hq, dnxhr_hqx, dnxhr_444) and Apple ProRes profiles (prores_proxy through prores_4444) with automatic pixel format mapping (yuv422p, yuv422p10le).
• Dynamic Video Format Querying (getAvailableVideoFormats): Added LRecorder::getAvailableVideoFormats(container, codec) and C API now2sdk_recorder_get_available_video_formats. Prevents UI option mismatches by constraining available resolution/FPS selections based on container and codec limitations (e.g. strict VC-3 standards for dnxhd, SD PAL/NTSC for dv and mxf_d10).
• Multi-Channel Audio Recording Support (Up to 8-Channel / 16-Channel): Upgraded LRecorder audio pipeline to dynamically handle multi-channel layouts (such as 8-channel discrete broadcast PCM or surround sound), removing the previous 2-channel stereo limitation.
• LMultiview GPU-Only Engine: Redesigned LMultiview rendering pipeline to run 100% on GPU shaders. Removed all CPU scaling (sws_scale), offscreen intermediate cell framebuffers, and cell RGB textures. Frames are uploaded natively and scaled directly to target viewports on the main FBO in a single pass.
• Lock-Free Multiview Rendering Loop: Decoupled state locks from OpenGL uploads and draw calls in LMultiview, eliminating stutter during source switching and switcher transitions.

Version 0.1.0 (July 4, 2026)

New Features & APIs

• Modular UI Integration Plugins: Split Qt6 and GTK4 graphics backends out of the core libnow2sdk.so library into separate runtime-loaded plugins (libnow2sdk_qt.so and libnow2sdk_gtk.so). The core SDK now has zero compile-time and startup runtime dependencies on GTK4 or Qt6 widget packages. LPreview dynamically and transparently loads the correct backend plugin via dlopen/dlsym at runtime when the ui_framework property is set.
• LMultiview Class: Added high-performance multi-view monitoring grid compositor. Layouts can be configured with Program (PGM) and Preview (PVW) screens at the top, bottom, left, or right, with up to 12 grid camera inputs. Built-in tally borders (Red/Green), dynamic cell name labels, and live VU audio peak level meters per cell rendered on offscreen GPU context.
• PNG Watermark Logo Overlay: Replaced the old pixelated bit text watermark with a high-quality, pre-loaded transparent PNG logo (logo-now2media.png) scaled to half-size. Set default opacity to 30% (70% transparent) and added flipVertical support inside LLicenseManager::applyWatermark to handle vertically inverted OpenGL frames.

Performance & CPU Optimizations

• Audio Resampler Bypass: Bypassed CPU-based audio resampling (swr_convert) inside LLive::processAudioFrameInternal when target audio format properties are not explicitly set (defaulting to pass-through), dramatically reducing CPU overhead for monitored sources.
• Zero-Copy NDI Audio & Video Ingestion: Implemented custom av_buffer_create memory wrappers with dedicated free callbacks (ndi_frame_free_cb / ndi_audio_frame_free_cb) for NDI video and audio streams, avoiding heap allocations and memory copy overhead on the CPU.
• CPU Spin Locks Prevention: Added millisecond safety sleeps inside NDI and VTP slicer loops when no source frames are received, avoiding tight CPU spinlocks when sources are idle.
• NDI Bandwidth & Decoding Optimization: Set NDI color format to NDIlib_recv_color_format_fastest to optimize chroma upsampling during decoding.

Version 0.0.2 (June 28, 2026)

New Features & APIs

• LAudioMixer Class: Added a multi-channel audio mixing and VU level monitoring engine. Supports custom channel gains, channel mutes, master volume controls, and real-time left/right channel peak meters.
• Chain Pull-Rendering (Master Clock Sync): Integrated an on-demand frame pull mechanism. Switcher and Mixer layers can now synchronize their rendering pipelines to a downstream master clock, eliminating thread drift and display judder.
• Dynamic Timing Fallback Lifecycle: Implemented automatic thread control. The switcher thread automatically halts and joins when pull consumers are connected, and respawns a high-precision 50 FPS internal timer thread when all consumers disconnect.
• Hardware Ingest Drop Detection: Added automatic frame drop tracking for DeckLink inputs using the GetStreamTime() API, logging warnings ([DECKLINK DROP DETECTED] X frames dropped!) to console.

Reliability & Sync Fixes

• Jitter-Free DeckLink Audio-Video Sync: Video frame items and audio packets are now stamped with a unified steady_clock timestamp at the hardware callback level (VideoInputFrameArrived), resolving OS thread scheduling audio cuts and lip-sync drift.
• Limited Range YUV Correction: Fixed luma-key conversion values for broadcast inputs (mapping limited range [16, 235] to full range [0.0, 1.0]), preventing transparent slate/CG overlays from leaving grey halos or grass discoloration.
• NDI & VTP Dynamic Reconnection: Playout sinks (SwitcherSink / MixerSink) now dynamically re-allocate SwsContext and SwrContext buffers if NDI stream source dimensions, frame rates, or audio sample rates change mid-stream.
• Native GPU Planar YUV Ingestion: Added direct GPU texture uploads for YUV422P, YUVA422P, and YUVA420P formats, saving CPU cycles by skipping software conversions.
• Fast-Path SwScale Bypass: Bypasses sws_scale scaling pipelines inside LSwitcher if the source layer format and resolution already match target output standard, minimizing CPU copies.

Build & Licensing

• LGPL FFmpeg 8.1.2 Isolation: Compiled FFmpeg from source using LGPL-only flags. Resolved transition link configurations to exclude GPL dependencies (libx264, libx265) while maintaining Nvidia GPU hardware-accelerated NVENC.
• Cisco OpenH264 Integration: Integrated Cisco's patent-safe OpenH264 library as a CPU-fallback encoder for offline H.264 exports.

Version 0.0.1 (June 18, 2026)

• Initial Beta Release: Open-core C++ broadcast SDK for Linux featuring LFile playout, LLive capture, LRecorder encoder, LStream streaming sink, LCharacter graphics, and LMixer offscreen OpenGL compositor.