← slowmotionbirds.com  ·  field notes

The Footage
Levels.

Every stretch of slow motion bird footage gets graded 0 through 5. The grade decides where it goes: the daily clip, the films, the ambient mixes, or the archive. An automated pass assigns the first grade from how much is actually happening in frame; the final call is always human, and every human correction teaches the system this yard's version of interesting.

v1.0.3 · Updated Sun / 2026-09-20 / 9:22 PM CT

5banger

The 60-second clip you send people

Birds fighting mid-air. Hummingbirds doing hummingbird things. Multiple species on screen negotiating who owns the feeder. Peak motion, peak drama, no dead frames.

Feeds: daily social clip · film openers
4strong

Whole sequences that hold attention

Sustained interesting action: arrivals and takeoffs stacking up, active squabbles, several distinct moments per minute. Not quite a banger, never boring.

Feeds: the films · strong cuts
3the cut

Moving and going

Birds relocating, feeding flurries, real motion without aggression. The workhorse tier: most of what makes a watchable long-form film lives here.

Feeds: standard selects
2ambient

The background layer

Minimal activity, but deliberately kept. This is the connective tissue of a ten-hour ambient mix: mostly calm, so that when something moves in your periphery, it feels like you caught it yourself.

Feeds: ambient mixes · long-form TV loops
1logged

Bird present, nothing happening

A finch parked at a port, pecking, for a long stretch. On record, searchable, and almost never used.

Feeds: nothing by default · kept in the log
0dead

The camera was just running

Empty frame, swaying branches, no visitors. Flagged so nobody ever reviews it again.

Feeds: nothing · flagged dead

How grading works

A motion-analysis pass watches every frame and scores what the eye would: how much movement, how fast, how many birds, arrivals and exits, wing bursts versus idle pecking. Levels 0 through 3 come from that signal directly. The 4/5 line involves judgment (a fight and a feeding flurry can carry similar energy), so the top tiers start as educated guesses and get corrected in review.

The correction is the clever part: in the edit, each graded clip sits on the video track matching its level. Regrading a clip means dragging it up or down a track, which takes one second, and the system reads those corrections back and adjusts its thresholds for that camera setup. Reviewing the footage is training it.

Known edge case under calibration: the rotating feeder, whose spin reads as constant motion. Its zone gets its own raised baseline so the spin itself never inflates a grade.

Changelog

v1.0.3 (2026-09-20, 9:22 PM): Legibility pass, styling only. Nothing about the levels changed and no words on the page changed. Nothing is under 12px now (the word under each level number was 9px), paragraphs and the text inside each level card are 16px, and the grey used for labels, the two lowest level numbers and this changelog moved to a lighter sibling that measures 7.9 to 1 on the page and 7.5 to 1 on the cards (it was 6.2 and 5.9 against a floor of 7). The back link and footer links are full 44px touch targets, keyboard focus shows a visible ring, and smooth scrolling steps aside for anyone who asks their device for reduced motion.

v1.0.2 (2026-09-19, 9:52 AM): A quiet Privacy link in the footer, pointing at the plain-language notice for this website at coltonland.com/privacy (what the visitor counter counts, what is never done, how to reach Colton). Nothing about the levels changed.

v1.0.1 (2026-09-18, 5:29 PM): Added the site's visitor counter (Cloudflare Web Analytics: no cookies, counts visits, not people). Nothing on the page itself changed.

v1.0.0 (2026-08-08): First public write-up of the level system: six levels, definitions, destinations, and the track-based regrading loop.