DELIBERATE

Subjects sit an average of 10% from the power points — framing this disciplined across a dialogue two-hander is what a feature DP would deliver.

Frames stay clean and uncluttered — the eye goes to the actors, nothing competes for it.

Native 16:9 delivered as 16:9 landscape — no crop or misfit for the platform.

The floor sits near silence — a controlled-location recording with no audible hiss under the dialogue. This is the cleanest capture in the sample set.

Dead-centered image — no left/right imbalance.

A measured pace for scripted dialogue, and with speech filling 48% of the runtime the film breathes between lines. Read as narrative dialogue, not a presenter format — so it's reported, not scored against a presenter target.

One filler across 887 words — written and performed, exactly as scripted dialogue should read.

The long pauses are the film's pacing — held looks and beats between lines. For a drama the engine treats them as intent, not dead air, so they carry no penalty.

Three strong luminance transitions across the whole film — effectively zero per minute. No strobing or flash risk.

A dark image by choice — the engine reads the 20% mean luminance as a low-key mood, not underexposure, so it's reported rather than penalized.

Highlights held — only 0.5% of pixels clip, even against the bright classroom windows.

Nearly a third of pixels sit at black — consistent with the low-key grade, so it's flagged as a creative choice rather than a defect.

The image uses 71% of the available tonal spread. A touch more mid-tone separation would help the darkest scenes survive a phone screen in daylight.

Healthy contrast inside the low-key palette — the image is dark, not flat.

The 14% cool cast is read as part of the film's warm/cool grading scheme, not a white-balance error — informational only.

Restrained, naturalistic color — fitting the drama.

The grade shifts as scenes change — read as scene-driven grading on a narrative, not drift to be fixed.

No visible color noise, even in the deep shadows.

A measured, classical cutting rate over the 14:25 runtime — continuity editing in service of performance.

Right inside the 2–15 s film target — unhurried but never static. A vlog threshold would flag this; the film threshold doesn't.

Long held takes set against quick reaction cuts — high variety that keeps a slow base tempo alive.

A gentle cutting pulse, consistent with the two-hander format.

The opening 30 s cuts at the same pace as the body — for a short film that's a pass, not a slow start.

Most of the film sits in takes longer than 8 s. The format absorbs much of that, but the longest stretches coincide with the tension dips flagged under Narrative — a few reaction cuts would re-anchor attention.

The cut slows down across the film's thirds — partly the script's design as the confrontation lengthens, but it leaves the final act leaning entirely on performance.

Essentially none — clean continuity editing throughout.

60% of screen time on faces — exactly the shape a dialogue drama should have.

The hook is real — an intro card and a loaded first meeting — but the film stops rather than ends: no closing card or formal outro after the final scene.

Seven distinct sections read clearly in the signal — the escalation from reunion to confrontation to aftermath is well-segmented.

An intro card opens the film; nothing brands the ending.

The engine hears what is likely a spoken call-to-action near the end of the upload.

The hook and shot variety help, but long quiet stretches between re-hooks soften mid-film tension — the price of the patient build, and the one structural number worth working on.

At ~3 s between sampled frames, frame-to-frame motion measures different shots, not camera shake — so the engine reports it instead of scoring it, and reads stability from motion fluidity and the horizon.

Residual jitter consistent with handheld operating — part of the film's visual language, not penalized.

Camera speed changes read as uneven — some moves accelerate and settle abruptly. A light stabilization pass or smoother operating would polish the handheld style without erasing it.

Detected as shoulder-mounted handheld — fitting for an intimate drama.

Level throughout — no persistent tilt.

The subject reads sharp in 57% of usable shots — and the dark frames are already excluded from the call, so the low-key grade isn't being misread as softness. Shallow depth of field explains some of it; the rest are genuine misses.

The single biggest image issue: on a two-hander built on close-ups, eye-level focus is everything. Lock focus on the speaking actor or cut the softest takes.

Clean — no visible ISO grain, even in the dark scenes.

Blocking shows on the 8×8 grid — the direct cost of the 0.3 Mbps encode, most visible in the flat, dark areas this film lives in. Fixing the bitrate (Priority 2) fixes this.

High-frequency detail confirms true 1080p — not an upscale.

Text rides on nearly every shot — chiefly the persistent channel watermark, plus the intro card and credits. 153 overlays were read and measured in all.

White lettering over the bright classroom backgrounds is where it ghosts out — 59% of samples fall under the 4.5:1 readability line. A stroke, shadow or plate fixes every instance at once.

Short, comfortable lines — never a wall of text.

Standard and fully compliant — and 24 fps is the right cadence for the format.

AV1 is the right codec, but 0.3 Mbps starves a dark, gradient-heavy image — it's the source of the blocking measured under Sharpness. Re-export at a healthier bitrate.

Mastered far under the −14 LUFS reference. Platforms only turn loud videos down — they won't bring a quiet one up — so the film plays much softer than everything around it.

Peaks sit far below the −1 dBTP ceiling — which also means there is plenty of clean headroom to raise the master.

Very wide macro-dynamics — cinematic in a theater, punishing on a phone speaker, where the whispered passages disappear under a master this quiet.

Zero clipped samples — the recording chain is clean.

Loud and quiet passages swing far apart. Ride the dialogue bus or apply gentle program compression so the quietest lines survive normalization.