Lucid

KEYBOARD SHORTCUTS

ENHANCE TAB

The SMART AUTO, AUTO LEVELS, AUTO COLOR, and LOW LIGHT presets are adaptive: they analyze the image histogram and channel means and compute exposure, contrast, and white-balance values from what they find — the computed values appear in the sliders, so every automatic decision is visible and adjustable. Static presets (FORENSIC, DOCUMENT, CLARITY, VIVID, MUTED) cover common looks. The Advanced section contains a proper unsharp mask (amount, radius, threshold), gamma, and color balance. Filters include CLAHE, auto white balance, and dehaze. A dot on the tab indicates non-default settings.

TRANSFORM TAB

Zoom, rotate, flip, and crop your image. Use Crop Mode to select a region, then Apply Crop (reset rotation/flip first — crop works on the unrotated image). Perspective correction lets you click 4 corners of a document, screen, or plate photographed at an angle and warp it flat. ML Super-Resolution neurally upscales the image entirely in-browser (3x built-in model, or load your own ESRGAN-style .onnx). TEXT & PLATES mode adds a forensic pipeline around the model: denoise, iterative back-projection (the result is forced to stay consistent with the true source pixels, suppressing hallucinated strokes), and stroke-tuned sharpening. For plates: crop, perspective-correct, then upscale in TEXT & PLATES mode — use DOUBLE passes for very small crops. For maximum quality, the REAL-ESRGAN X4PLUS option downloads a state-of-the-art model (~67 MB, one time, SHA-256 verified against the published hash) directly through Lucid; SAVE MODEL TO DISK keeps a copy for fully offline reuse via CUSTOM .ONNX. The URL is pinned to an immutable repository revision and is editable in case you prefer a mirror.

ANNOTATE TAB

Draw shapes (rectangles, circles, arrows, lines), add text labels, measure distances, or blur sensitive areas. Watermark and Caption tools add overlays. All actions are undoable with Ctrl+Z in the order they were made.

OSINT TAB

Reverse image search, in-browser OCR (whole image or a drag-selected region — region mode is far more accurate on plates and signs), and Error Level Analysis (ELA) — a quick manipulation screen that highlights regions whose JPEG error level differs from their surroundings.

METADATA TAB

EXIF data with copy buttons, GPS map, raw tag dump, JPEG quantization tables with estimated encoder quality, embedded thumbnail comparison (mismatches betray editing), SHA-256 hash, and a session history audit trail. EXPORT CASE REPORT produces a self-contained HTML report — hashes, images, audit trail, settings, metadata — archivable as-is and printable to PDF. Lucid also installs as an offline app: after one online session, everything (including downloaded models) works with the network unplugged.