Changelog

Release notes for Videre. All notable user-facing changes are listed here.

1.0.0

Platforms

• NVIDIA Jetson Orin (Nano, NX, AGX) is now supported, alongside x86 and Raspberry Pi 5. Single binary, same browser UI.

New tiers

Multi-camera workspace is exclusive to Pro Multi and Subscription Multi. Community, Pro, and Subscription remain single-camera.

Added

• Jetson Argus ISP panel with live-apply controls: auto-exposure (lock, EV compensation, exposure time, analog gain), AWB (10 presets, lock, manual R/B gains), denoise mode + strength, edge enhance mode + strength, color saturation, and Frame Timing (cap frame duration; label turns orange when throttled below max fps). Advanced subset (anti-banding, focus/aperture position, ISP digital gain, saturation bias) under Advanced Device Controls. Values survive stop/start cycles.
• Max Performance mode (Jetson) at the top of Setup. Bypasses filter chain, downscaler, and histogram, feeding raw NV12 straight into the HW JPEG encoder for native-resolution preview at peak fps. Snapshot/recording prompt to disable it first, with a "Don't ask again for this session" option.
• WebRTC streaming for low-latency browser-native preview, bundled in Pro and Subscription release packages on Jetson and x86 (libdatachannel runtime, third-party notices included). NVENC (Jetson hardware encoder) support is experimental and will fall back to software encoding.
• Multi-session workspace (Pro Multi / Subscription Multi). Built-in layouts: single, split-vertical, split-horizontal, 2x2, custom. Drag session pills into tiles; divider snap guides; per-feed metadata badges (fps, resolution, pixel format, session/camera name); linked or per-view zoom; right-click drag pan per viewport; geometry and assignments persist across reloads.
• Multi-target snapshot: focused session, all visible sessions, custom selection, or composite workspace JPEG (browser-side stitched export).
• Multi-target recording: focused, all visible, or custom-selection backend recording. One file per session.
• Overlay system with a dedicated Overlays section and per-session state. Geometric overlays: Crosshair, Center Lines, 3x3 Grid, Aspect Frame, Scale Bar, Ruler. Analysis aids (preview-only): Zebra Exposure, False-Color Clipping, Focus Peaking. Per-overlay enable, rename, opacity, color, line thickness, stack order, position. Presets for Alignment, Targeting, Framing, Measurement. Paid tiers can save named local presets, enable recording burn-in, and upload custom SVG overlays (preview-only for now).
• OCR detector (EAST text-region detection + Tesseract) on x86, Jetson, and Raspberry Pi 5. Runs alongside shape and barcode detectors.
• Image mode RAW support (Pro / Subscription). Load camera RAW files via LibRaw with three render modes: Camera, Raw, Filmic. Adjustable tone-mapping parameters.
• Camera Panel reorganisation. Sections: Setup, Capture, Image Pipeline, Processing, Detectors, Advanced Device Controls. Layout is config-driven and can be overridden with ~/.config/videre/right-panel.json. Low-value sensor internals moved out of the default view.
• Setup panel reporting: Active Backend (e.g. v4l2, jetson-argus), Preview Path (passthrough, native JPEG, conversion/downscale fallback), Output Resolution, and Pixel Format selector for RPi5 PiSP (NV12 default, RGB3 selectable when stopped).
• Diagnostics page (/diagnostics, linked from Settings) now reports active backend, native pixel format, performance-mode state, and fallback reasons in addition to existing hardware info, USB bandwidth utilisation, VA-API availability, and live pipeline timing breakdown per stage.

Changed

• Activation credits: Pro and Pro Multi grant 3 activation credits (lifetime). Subscription and Subscription Multi grant 5 credits per year; yearly renewals add another 5 to the same license.
• Cached activation tokens are preserved across rebuilds and upgrades on the same machine until backend validation invalidates them. Startup no longer discards cached state purely because of local fingerprint drift.
• Raspberry Pi 5 PiSP ISP now defaults to NV12 output for the WebRTC and native JPEG fast paths. RGB24 (RGB3) remains selectable from Setup → Pixel Format while not streaming.
• Output Resolution model: the setup panel presents delivered stream sizing as Output Resolution. When the active backend ties output size to the chosen sensor mode, the UI states that explicitly.

Fixed / improved

• Jetson UVC MJPEG analysis: enabling filters or analysis overlays on Jetson now uses guarded software JPEG paths instead of NVIDIA's libjpeg-compatible stack, avoiding hard crashes on some UVC MJPEG streams. Native NV12/I420 preview path still uses the hardware encoder.
• IMX708 ISP streaming on RPi5 (embedded pad EPIPE issue resolved).
• RPi5 sensor mode switching without closing and reopening the camera.

Known limitations

• SVG overlay snapshot/recording burn-in: not yet implemented. SVG overlays render in the live preview only.
• OCR runtime data: requires Tesseract language data (e.g. tesseract-ocr-eng) and the EAST model file. If either is missing, OCR stays unavailable while the rest of the app continues.

0.9.2

Platforms

• Linux x86-64 (.deb, .AppImage, .tar.gz)
• Raspberry Pi 5, arm64 (.deb, .tar.gz)

Tested camera modules

• Raspberry Pi Camera Module v1 (OV5647), v2 (IMX219), v3 (IMX708)
• Raspberry Pi HQ Camera (IMX477)
• Any V4L2-compatible USB camera

Highlights

• Single binary, browser UI, no Qt / Electron / desktop environment required.
• Live JPEG streaming over WebSocket. H.264 decoding via FFmpeg.
• Live histogram (R/G/B/luma) with log/linear scale, clipping indicators, and 5-second analyze mode (exposure, color balance, flicker DFT).
• ROI analysis: per-channel min/max/mean/std-dev plus horizontal/vertical line profiles.
• Image processing pipeline: Gaussian blur, median blur, sharpen, grayscale, binary threshold, Canny edges.
• Shape detectors (Contour, Circle, Blob) running concurrently on separate CPU cores.
• Barcode / QR scanner via ZBar.
• Snapshot (JPEG, lossless TIFF, raw V4L2 bytes) and AVI recording.
• CLI commands: videre list, info, capture, record.
• Raspberry Pi 5 PiSP ISP support: tuning file selection, AWB, black level correction, lens shading, defect-pixel correction, denoise, sharpen, manual lens focus, one-shot autofocus.
• Bayer demosaic for 8/10/12-bit packed and unpacked CSI-2 formats.
• Hardware crop vs. UVC ROI auto-detection per device.
• Performance diagnostics (/diagnostics): hardware info, USB bandwidth, VA-API availability, actionable recommendations.
• Tiers: Community (free, mandatory usage data), Pro (€89 one-time), Subscription (€12/mo), Source License (€3,999).