One video in, ready-to-post YouTube thumbnails out. The app
transcribes the video, mines it for usable stills of Justin, has a strategist
draft ranked thumbnail concepts, and renders each one with Justin's real
likeness — which you then refine until one is worth posting.
The flow
- Pick the video type at the top — it sets the whole look (see below).
- Drop in the video or audio file. It's transcribed automatically
with speaker labels, and videos are also mined for stills of Justin (and any second
person), each labeled by expression. No file? Open the Transcript box and paste one instead.
- Add creative direction (optional) — "dark background", "heavy, sad
energy", "make Justin's face extra big" — it steers every concept and image. Change it
after drafting and a one-click Re-draft button appears.
- Generate concepts. The strategist reads the transcript and returns
ranked concepts, each with on-image text, a layout, props, candidate titles, and an
image prompt (all visible under the card's Details).
- Generate images card by card. Drafts render cheaply so you can
explore freely; each card collects its takes on a small carousel.
- Refine and finish — edit, re-imagine, then finalize the winner.
- Write the description (panel below the cards) — one click turns the
run's transcript into a ready-to-paste YouTube description: a hook line above the fold,
a plain-language summary, chapters anchored to the video's real timestamps, and the
channel's subscribe links. Edit it in place, Copy, and paste into YouTube Studio.
The four video types
- Platypus Economics — the standard solo look for native channel videos.
The theme pills under it restyle the run (pick one or several — each concept renders
once per theme). Themes lock in when concepts draft; picking a different theme on a
finished run offers Start a new run — same video, which reuses the transcript,
mined stills, and title with no re-upload.
- TV appearance — the thumbnail becomes the broadcast playing on a
classic TV set. The network is auto-detected from the video (or pick it in the
Network dropdown) and its real logo is composited onto the cabinet — you can change
the network mid-run and future images follow.
- Cohost / Interview — two faces, Justin always on the left. Declare
the cast before uploading so frame-mining keeps shots of the second person.
A cohost is a recurring person saved in the reference library (Settings → People);
a one-off guest is rendered entirely from the video.
- Prior format — the earlier channel look: Justin in his room with the
boxed caption stack, captions typeset in the real brand font. Videos that never show
the room can use one of Justin's studio backdrops via the Backdrop dropdown.
- Stark — a minimal poster: one flat solid-color background, a big stark
headline on the left (1–4 words) with a single word or character in a neon accent,
and Justin on the right as a tight face-forward portrait carrying the emotion. The
headline is typeset in the real brand font and is live-editable (accent, background,
and which word is colored) with no re-render.
Working a card
- Generate image — renders a draft. Every image is identity-anchored:
Justin's canonical reference photos plus an expression-matched still from this video
ride along, so it's really his face. Click a still on the card's strip to anchor a
different pose. After the first draft the button becomes Redraw — a fresh
take on the same concept.
- Edit — describe one change ("make the price tag bigger") and
only that changes. Limit to area lets you drag a box so everything outside it is
protected. Every version lands on the carousel — nothing is overwritten, and an edit
builds from whichever version you're viewing.
- Re-imagine — keeps the concept's idea, hook and titles but invents a
fresh visual as a new sibling card; Instruct lets you steer it first.
- Download re-renders the take you're viewing at high quality
for extra detail (faithful to this take's composition), then downloads it upscaled to
4K. Takes that are already high quality download as-is, with no re-render.
Keyboard shortcuts
On the grid, the current card is the one you last hovered or arrowed to:
- ← → — page through the current card's takes;
Home / End jump to the first / latest.
- ↑ ↓ — move between concept cards.
- Enter — enlarge the current card; E — open its Edit box;
Esc — drop the selection.
- ? — open this guide.
Click any image (or press Enter) to enlarge it: ← →
step through its versions, ↑ ↓ jump between concepts,
Esc closes. Ctrl+Enter applies an edit.
Nothing is lost
Every run saves continuously — transcript, stills, concepts, and every image version —
and reloading the page restores it. Open (top bar) lists past runs;
New starts a blank one.
Settings (⚙) and cost
The cog manages the reference photo libraries (Justin and cohosts), known speakers'
voice samples for transcription, the models, every prompt the app sends, and the knobs:
concepts per run, draft takes per card, and draft quality.
Images render only on demand, so you control spend: a draft or an edit costs about a
cent, finalizing at high quality about twenty cents, and concepts a fraction of a cent.
