Rule of Thirds
Place key elements on the intersections of a 3×3 grid. The composition classic.
Score your photos 100% on your device. Real-time live scoring and deep static analysis — 63 rules, 17 ML models, works offline. Every photo becomes your best one.
Beyond real-time AI scoring, Compoz offers AI-powered tools to perfect every shot.
Enable Auto-Shoot mode and let Compoz automatically capture when your composition exceeds your chosen threshold. Perfect for street or nature photography.
Import a reference photo and trace its composition in real time using a hand-drawn pencil-sketch contour overlay on the viewfinder.
Rate your photos in review mode to teach Compoz your style. The algorithm adjusts the 63 rule weights to your preferences — your scores become unique.
Progress with daily and weekly challenges: portraits, landscapes, excellence scores. Keep your streak and develop your eye.
Each rule is evaluated in real time and adapted to the detected scene type
Place key elements on the intersections of a 3×3 grid. The composition classic.
Harmonious proportion (~1.618) that guides subject placement for natural balance.
The subject follows the logarithmic curve, drawing the eye in a fluid motion.
Detects visual balance between image halves, ideal for architecture and reflections.
Measures space around the subject. Too tight or too empty — the engine detects both.
Uses natural elements (doors, arches, branches) to frame the subject.
Checks the space between the top of the head and the frame edge in portraits.
Measures the real headroom from the top of the hair, not the forehead. The pro framing band sits between 8 and 15% of the frame height.
Positions eyes on the upper third for a natural, engaging portrait.
Detects lines (roads, rails, walls) that guide the eye toward the subject.
Measures camera tilt from detected horizontal lines. A tilted horizon throws the photo off balance.
Detects objects visually merging with the subject's head — like poles or trees that appear to grow from behind. Checks above and both sides.
Analyzes whether the zoom level matches the scene type (portrait, landscape, macro).
Evaluates whether you're at the right distance from your subject. Suggests getting closer or stepping back for optimal framing.
Evaluates light distribution via the 32×32 brightness grid to detect dark zones.
Detects backlighting conditions and guides you on how to use or reposition the light source creatively.
Meters the face skin for blown highlights or crushed shadows. Tolerates intentional Rembrandt-style shadow when the lighting is directional.
Measures the dynamic brightness range across the image. Good tonal contrast adds depth and dimension.
Analyzes color temperature consistency. Detects golden hour and saturation for harmonious tones.
Checks that joints (knees, elbows, ankles) are not cut off by the frame edges.
Evaluates the proportion of sky in the image. Ideal is 20-45% for a balanced landscape.
Analyzes depth layers (foreground, midground, background) for a three-dimensional image.
Estimates gaze direction from detected eyes. A camera-facing gaze creates connection with the viewer.
Evaluates background clutter: brightness variance, depth separation, and distracting object count.
Ensures enough space in front of the subject's face direction. Prevents the subject from looking into the frame edge.
Estimates overall image sharpness using Laplacian edge detection on the 32×32 brightness grid.
Detects repeating visual elements (columns, windows, waves) using autocorrelation. Rhythmic patterns add compositional strength.
Analyzes the brightness histogram for blown highlights and crushed shadows. A well-exposed image uses the full tonal range.
Checks whether the phone orientation (portrait/landscape) matches the scene type for optimal framing.
Evaluates subject placement along the frame's diagonal lines, creating dynamic tension and visual flow.
Compares the brightness center of mass with subject position. Aligned visual weight creates natural, balanced compositions.
Evaluates spacing, size consistency, and alignment of multiple faces in group photos for balanced composition.
Checks for interesting foreground elements that create depth. Rocks, flowers, or paths in the lower frame add visual layers.
Measures how well the subject stands out from the background using depth, brightness contrast, and color temperature differences.
Detects directional motion blur by comparing horizontal and vertical gradient energy. Sharp subjects with clean backgrounds score high.
Evaluates color temperature consistency. Neutral tones score high; mixed lighting or strong color casts are penalized (golden hour excepted).
Detects camera shake, missed focus, and motion blur to ensure sharp photos.
Flags overly busy compositions with too many competing elements.
Checks if your composition matches the lens used — wide angle, normal, or telephoto.
Scores whether the horizon line sits on a thirds or golden ratio line rather than splitting the frame in half.
Detects bright spots or subjects touching frame edges that pull the viewer's eye out of the composition.
Every composition rule, every ML model, every score — runs locally on your phone. Works offline. AI Coach is an optional research beta that helps us improve the on-device model; once our model is ready, the beta folds back into live scoring and disappears.
VisionCamera captures the camera feed at ~2 FPS
ML Kit extracts faces, labels and a 32×32 brightness grid
FrameAnalyzer builds features: subjects, lines, scene
63 rules evaluated, weighted by scene → score 0-100
ML Kit Face Detection locates eye positions, bounding boxes and facial landmarks. This data feeds the Headroom and Eye Line rules.
The BrightnessExtractor samples YUV data into 1,024 cells. The LineDetector (Threshold + Sobel) detects leading lines and the horizon.
13 scene types detected (portrait, landscape, food, architecture…). Each scene activates a subset of the 63 rules with different weights.
Object detection, pose estimation, selfie segmentation, NIMA aesthetic scoring, MiDaS depth estimation, plus Apple Vision (face capture quality, person segmentation, body pose, animals, horizon, foreground instances) — orchestrated by an intelligent per-scene scheduler running on the Apple Neural Engine.
On iPhone Pro and iPad Pro, Compoz taps the LiDAR sensor to score how well your subject separates from the backdrop in real depth — not estimated. The HERO_BACKDROP_DEPTH rule rewards a clean ≥2 m gap and flags subjects glued to the wall behind them. Other devices fall back to MiDaS on-device depth estimation, no cloud.
Compoz uses your GPS position and the time of day to calculate the sun's exact position. During golden hour, the app adjusts scoring to reward warm backlighting, dramatic skies, and silhouettes — and tells you exactly how many minutes of magic light remain.
The built-in compass detects if you're shooting into the sun or with the sun behind you. During golden hour, backlighting becomes a creative tool — Compoz rewards it instead of penalizing it.
Real-time coaching is free. Unlock Pro for AI Coach and progression tracking.
Everything people ask before downloading.
A live photo coach in your viewfinder: it analyzes your composition in real time and tells you what to fix BEFORE you press the shutter. No more "hope it came out well" — the photo is right the moment you take it.
No. All live analysis runs on the iPhone (Apple Neural Engine) and works offline. Only the optional AI Coach (Pro) sends the single photo you choose for an in-depth critique — it is never stored without your explicit consent.
The core app is free: live coaching, scoring on 63 rules, scene detection, gallery. Pro ($4.99/month or $39.99/year, −33%) adds the AI Coach, Reproduce This Shot and the full lessons.
iPhone (iOS) for now. An Android version is coming later — we're perfecting the app on iOS first.
Never. Compoz doesn't generate or alter a single pixel — it helps you capture the real photo, genuinely well taken. That's our red line: recreating a photo with AI is cheating.
Every score is linked to evidence: the 63 rules evaluated, what weighed in and why, plus a percentile computed against YOUR own photos. You can open "Why this score" on any shot.
Yes — that's exactly who it's built for. Advice comes as simple actions ("step back", "lower the horizon") and the Learn tab explains every rule with examples.
Download Compoz for free and start improving your photos today.
How it works
Before, during and after every shot — AI coaches you at each step
AI guides your composition
Open Compoz and point the camera. AI-powered composition grids, directional arrows and coaching tips overlay your viewfinder in real-time. 17 AI models detect the scene — portrait, landscape, architecture, food — and adapt their advice.
AI captures the perfect moment
An AI-driven live score from 0 to 100 updates as you frame. Set a threshold and auto-shoot captures the best compositions automatically. Found a shot you love? Use Copy Shot to reproduce its composition.
Review, learn, improve
Every photo you capture lands in your gallery with a detailed per-rule breakdown and concrete tips on what to work on next. Rate your favourite shots 1–5 and Compoz quietly tunes its 63-rule weights to your own style. For the most demanding photographers, our AI Coach research beta (opt-in) adds an extra layer of expert feedback on any photo.