Composition and focus
Review the rule of thirds, visual-weight distribution and an editable focal anchor. Interpret guides in the context of your scene.
Inspect framing, tonal contrast, color and texture. Add creative context, compare frames and create production-ready shot notes—without uploading your footage.
Import a frame, inspect the evidence, and add your own artistic interpretation.
No source loaded
Up to 12 PNG/JPEG/WebP images at once, or one video
Private by default · 220 MB per fileClick the frame to position your focal point. The orange dot is an edge-based texture centroid, not a detected subject.
Import multiple images or collect up to 12 editable frames.
Captured frames will appear here.
Pixel statistics and computational guides, not scene-semantic facts.
| Measurement | Value | Interpretation limits |
|---|---|---|
| Add a frame to view measurements. | ||
Add another frame to review adjacent-shot changes.
Analyze a frame to see context-sensitive review prompts.
Review the rule of thirds, visual-weight distribution and an editable focal anchor. Interpret guides in the context of your scene.
Compare luminance, highlights, shadows, color bias and brightness asymmetry without pretending to recover physical lights.
Build a sequence from still images or captured video frames, and export editable scene notes with measurable evidence.
Not reliably from a single arbitrary frame. The tool measures tonal gradients, edge texture and composition proxies. Camera angle, physical light direction, depth and storytelling intent are explicitly editable human interpretations.
Yes. Model Setup offers an explicit opt-in download of a free MediaPipe object detector. Detect subjects then uses that loaded model. Automated color and composition metrics work offline without the model. Live model integration requires internet access and browser verification.
No footage or images are uploaded by the application. On-demand AI inference uses locally processed frames but fetches third-party model/runtime assets and MediaPipe may collect runtime metrics.
Measurements are approximate sRGB pixel statistics, not calibrated HDR, RAW or camera-lighting measurements. There is no depth reconstruction, shot reverse-engineering or story understanding without additional validated models and context.
Preparing frame…
Review a sequence of stills or captured video frames with pixel-derived color, contrast and clipping measurements, then add your own focal and story annotations. Optional object proposals are a separate model-assisted step.
Example: Compare a warm and a cool setup from the same scene, mark the intended focal point in each frame, and export a contact sheet with your notes for the crew.
Limits and privacy: Metrics describe compressed image pixels; they cannot determine camera angle, scene depth, incident lighting or artistic quality. Projects can embed up to twelve frame JPEGs and notes rather than the original video. Optional MediaPipe and EfficientDet files download from external hosts after an explicit action and may disclose the visitor IP; real model inference remains unverified.