Shorts Autopilot: daily faceless video plans

One message in, one shootable plan out

The bottleneck in daily short-form is not the editing. It is the blank page at the start: what is this one about, how does it open, what is on screen at second four, what do I type into the stock library, what goes in the caption. That planning step is what turns "one a day" into "three this month."

Send a topic. Get everything needed to record and publish one faceless vertical video — no camera, no face, no presenter. The full package arrives in the first reply, every time. Not a conversation, not an outline you then have to develop.

What comes back

1. Three hook options — each pairing the opening line with the frame that runs under it, because a strong line over a dull visual still loses the viewer.
2. A second-by-second shot list — start time, duration, on-screen description and narration for each shot, adding up to your target length.
3. TTS-ready narration — the same script repunctuated for a synthetic voice, with numerals and abbreviations written the way they should be read aloud.
4. Stock-footage search terms per shot — concrete phrases that return usable results in a free library, with a fallback for when the first search comes back thin.
5. Captions and tags for three platforms — TikTok, Instagram Reels and YouTube Shorts each get their own length, structure and tag set.
6. A cover-frame specification — including a working character ceiling for text that still reads at feed-grid size, not at full screen.
7. A retention-risk review — the specific seconds where viewers are most likely to leave, with the concrete fix for each, before you build anything.
8. A series note — which part of this video is the fixed format and which is the variable, plus the next three topics in the same shape.

Timing is estimated from speech, not character count

English narration runs near 150 words per minute, so a second carries roughly 2.5 words and a 30-second video holds about 75 words. Counting characters instead is the most common reason a plan that looked right renders four seconds long — the gap between a line of short common words and a line of long ones is close to a full second, and it compounds across eight shots. Every narration line carries its own estimated seconds so you can see where the length went.

Judge the result by view rate, not view count

Early view counts on short-form cluster together regardless of quality. The share of the video people actually watched is the signal. Every plan ships with the standard: 75% and above means repeat the structure, 50–75% means tighten the weakest shot, below 35% means rebuild rather than rewrite the hook.

Three operating rules are baked into every plan, because each one suppresses distribution without producing any visible error: never re-upload a video that is already up, keep one account to one subject, and never publish restricted or "only me."

Thirty requests is a month of daily uploads

Each request returns one complete plan. That is why the skill will not spend a request on a clarifying question when it could send a working package instead — if the topic is missing it asks once and answers itself with a best-guess plan in the same reply, so no request ever comes back empty.

Honest limits

It plans; it does not produce. No video is rendered, no voice synthesised, no footage fetched, no file delivered. It does not post, schedule or read your accounts, and needs no login of yours. It has no access to live trends, tag volumes or your analytics, and will say what to check rather than invent a number. It never predicts views, followers or revenue. Check the licence on any footage or music you use, and the disclosure rules that apply to you if a video is sponsored.

It declines, with the reason: buying engagement, bot or multi-account posting, re-uploading someone else's video, stuffing unrelated tags, evading enforcement, impersonating a real person or brand, and fabricating performance numbers.