Read It For Me · promo analysis

Measured from osint_framework_read_it_for_me.mp4: 1462 frames, 48.7 s, 1080x1920, 30 fps, H.264 365 kbps + AAC 163 kbps.

Measured facts

2hard scene cuts
59.1%speech coverage
0.30avg motion / frame
-8.9avg loudness dBFS
0.6%accent colour share

Scene breakdown

SceneSecShareMotionAccent %

Loudness per second

Seconds 3, 6, 12, 17, 25, 32 and 45 sit 6 to 10 dB below the rest. Those are the gaps between narration lines, where only the 55 Hz sub and the noise bed remain, so the mix never breathes.

What the numbers say about the product

Finding 01

It is a slideshow, not a demo

Two hard cuts in 48 seconds and motion of 0.16 to 0.62 per frame. Nothing is actually being investigated on screen. The product promise is "it runs the lookups", and the video never shows a lookup happening. The website must show the run, live, in the browser.

Finding 02

The tree is the only memorable visual

The tree scene carries the highest motion in the whole film (0.56 mean, 20.6 peak) and the hook and close, the two scenes with the highest edge density, are just big type. The interactive category tree is the brand asset. Put it on the homepage and let people click it.

Finding 03

Findings are the retention peak

The evidence-card scene is the longest block at 10.3 s and the loudest seconds land at 19, 29, 38 and 41, all inside the findings and confidence scenes. Cards with a value, a source and a timestamp are the thing to sell. Make the site output exactly those cards.

Finding 04

It is too dark and too flat in colour

Mean luminance 27 to 35 out of 255, about 11 to 14% of full white, and the top eight colours are all near-identical dark blues covering 33% of every frame. Cyan and amber together are under 3%. A wall of dark blue reads as generic terminal. The site needs one bright signal colour used only on hits.

Finding 05

The voice is the weakest asset

Mean zero-crossing rate 0.22, harsh and robotic, and speech fills 59% of the runtime with no room to breathe. Do not build the product story around a text-to-speech narrator. Build it around the output being read back, and let the user choose the voice.

The website to build from this

Recommended build

An interactive OSINT tree you can type a target into

Homepage is the framework's category tree, drawn and clickable. Type a username, email, domain, IP or image URL. Branches light up as they run, then collapse into finding cards with source, timestamp and confidence, plus an explicit "not found" list. A read-aloud button speaks the findings, which is the one thing the video proved works. Everything runs client-side against a seeded demo dataset, so it needs no back end to feel real.

Secondary

A clip generator that fixes the video's own flaws

Take any finding set and export a 9:16 clip: brighter palette, one accent colour on hits, a real typing animation, and a cut every 2 to 3 seconds instead of every 24. The analysis gives the exact spec to beat.