Pull a clickable cover from the finished clip | AI Video Factory

The video is ready to post. The cover is still a random grab in the editor: blur, or a face cropped out, and clicks drop first. A calmer path in the studio: pull keyframes, score sharpness and optional faces, lay a title, then export the platform size. No second design project.

Decide if the frame can be a cover

Talking-head, explainer and tutorial clips with a person, a prop or a contrast work best. Opening logos, caption-only cards and near-black slates should not win, no matter the score. If you have not shot yet, draft a script or captions first, then come back.

After upload, wait for detection, then tick Smart cover. Vertical files lean 9:16, landscape 16:9; keep 1:1 for feeds that ask for a square.

From finished file to cover

1. Open cover in the studio and drop in a local MP4 or MOV. The file stays in the browser.
2. Wait for keyframes. Sharp frames, optionally boosted by a face, rise to the top. Do not lock the first tile — skim the first few.
3. Type a title, or let the model write an 8–16 character hook. Keep type off faces and key props. Templates change decoration, not the crop.
4. Export PNG per platform: 16:9 for YouTube / Bilibili, 9:16 for Shorts / Reels, 1:1 for some feeds.

The free plan includes one cover run per month. Ratios, titles and quota are on Features and Pricing.

Lock only three things

• Face and subject: after the crop, eyes or the product stay in frame. Drop half-faces.
• Length: a cover line should scan. Past ~16 Chinese characters or a short English clause, cut — do not paste the whole voiceover.
• Contrast: if type melts into the background, change template or frame. Do not only max the font size.

What usually breaks

Treating the cover as “any frame plus bold type”: blowouts, motion blur and the back of a head get a click and a swipe-away. Let the model rank sharpness; you own faces and copy. Three platforms from one file: change the ratio and export again — do not rerun keyframes.

The full post order is in creator post in 2026. For voice or translation on the same clock, see AI dubbing and bilingual captions.