YT Clipper - AI Content Studio
A free Windows desktop app that turns long videos into captioned shorts, and a short topic into quote images and
narrated videos, then posts or schedules all of it to your social accounts. It runs on your own PC: no account, no
cloud, no subscription. Your keys and your content stay on your machine.
Download: ytclipper.xyz · Built by Hasbiyallahu Jafaru
What it makes
Clips from a video. Paste a YouTube link or pick a file on your computer. The app transcribes every word, finds
the moments that stand on their own, and cuts each one into a vertical short of at least a minute. Dead air and
filler words come out, the crop follows the speaker's face, the strongest line can play first as a cold open, and
captions are burned in in the style you pick. Every clip comes with its own title, hook and post copy for each
network.
Quote images. Give it a topic and a genre, or bring your own quotes (typed, pasted, or imported from a CSV, with
a sample file and a ready-made AI prompt for writing one). Every quote is a real, documented line with its author.
Each image is laid out in one of 13 typographic designs (noir film still, paper note, Swiss poster, risograph print,
instant photo and more) over a stock photo picked to match.
Narrated videos. The same brief becomes a spoken short: a script written hook-first, read by a Gemini voice you
choose, cut across several stock shots with the genre's colour grade, film grain and optional karaoke subtitles.
Before anything full-size downloads, you see the footage the app proposes as thumbnails and keep or swap each pick.
Download and render then fetches exactly those and renders every piece on one progress bar, with a step-by-step
log that shows which piece and which step is running, finished or failed.
Posting. Connect your accounts once through Zernio (free for the first two). Post a single
clip or image, or press Schedule all: choose the first post time and a spacing (every 10 minutes, 15 minutes,
hour, day or other day), adjust any single time, and send the whole batch. Each item then shows Scheduled, Posted or
Failed. Or export a folder with the files and per-platform captions and upload them yourself.
Genres
Content is organised by format rather than mood, because that is how the platforms reward it. Each genre has its own
quote voices, narration style and edit grammar.
| Genre | Quotes from | Cut |
|---|
| Hope | Keller, Dickinson, Whitman, Hugo, Emerson, Tagore | 2.6-3.2 s shots, dissolves, warm lift and halation |
| Hard Truth | Roosevelt, Douglass, Seneca, Epictetus, Nietzsche | 1.4-1.9 s shots, hard cuts, bleach bypass, coarse grain |
| Stoic Wisdom | Marcus Aurelius, Seneca, Epictetus, Musonius Rufus | 3.0-4.0 s holds, mostly locked off, cold and desaturated |
| Historical Voices | Documented words of public-domain figures | 2.4-3.0 s, black beats between shots, sepia, heavy grain |
| Book Wisdom | Lines from public-domain classics | 2.6-3.2 s, alternating dissolves, warm paper mids |
| Cinematic Minimal | Lao Tzu, Basho, Rumi, Thoreau, Muir, Rilke | 3.5-4.5 s, fade to black, teal-orange, letterbox, one speed ramp |
Quotes are never invented: every one is a real public-domain line (published before 1929) with its author named.
How it edits
Renders are meant to look edited, not illustrated. A narrated piece cuts between several distinct shots (different
subject, different place) on the breaths in the narration, using real word timings. Stills get a slow push, every
join is a proper transition, and the whole timeline gets one grade so unrelated stock footage reads as one film.
Music ducks under the voice. A render that opens on a black frame fails its own quality check. The director's spec is
in .claude/edit-grammar.md.
Status
Version 0.1.0 is released and downloadable from ytclipper.xyz (Windows, about 256 MB,
unsigned for now, so Windows SmartScreen asks once on first run).
The pipeline is covered by 224 backend tests and an end-to-end smoke run that drives the real app against local
fake services (apps/desktop/scripts/smoke.mjs), plus a check of the packaged build. Real-key runs (OpenRouter,
Pexels, Zernio) are verified in part; NEXT_STEPS.md lists exactly what has and has not been tried
for real.
여기까지가 README 앞부분입니다. 나머지는 원문에서 읽을 수 있습니다. README 전문 보기 ↗
