You’re being designed (scroll → log → steer)

Template (five equal segments; list exact seconds for the chosen duration, e.g. 10 → 2s each, 15 → 3s each):

• Segment 1 — YOU THINK YOU’RE SCROLLING (or YOU’RE NOT REALLY CHOOSING).
• Segment 2 — THE FEED IS SCROLLING YOU (or equivalent flip).
• Segment 3 — EVERY PAUSE GETS LOGGED.
• Segment 4 — LAYOUT NUDGES YOU.
• Segment 5 — THEY STEER WHAT YOU SEE.

User may swap wording but must keep five windows unless they explicitly compress (then show full custom schedule in confirmation). Default duration is 10 when unspecified.

Dependencies: WERYAI_API_KEY + Node.js 18+. {baseDir}/scripts/video_gen.js + WERYAI_VIDEO_API.md. No other Cursor skills required.

Prerequisites

• WERYAI_API_KEY must be set; Node 18+; https images only.
• Non-empty model.

Security, secrets, and API hosts

• WERYAI_API_KEY: Secret; requires.env / primaryEnv.
• API hosts fixed in video_gen.js.

Prompt expansion (mandatory)

• Duration: Default 10; 5 or 15 when allowed and requested. For five equal beats, each caption span = duration / 5 seconds (state numeric ranges explicitly in the prompt).
• Timeline scaling: If merging to fewer than five lines, list new non-overlapping windows in the confirmation table.

Visuals: Thumb on glass, tiles exploding into data streams, recommendation grid, heatmap vibe, dark mode neon accents, glitch tastefully; hook in the first segment must be visually loud.

Typography: Bold sans, white + black outline, lower third, no overlap between the five lines.

Motion: Continuous—scroll inertia, parallax UI, particles.

Audio: Sub pulse, digital risers if generate_audio true.

Negatives (Kling): illegible HUD text, watermark, crowded captions.

Language: English on screen unless user requests otherwise.

Confirmation: Full expanded prompt with explicit non-overlapping second ranges for all five lines (derive from duration).

Niche checklist

• Reads as critique of systems, not attacks on named individuals.
• Avoid implying illegal surveillance claims unless user explicitly frames fiction.

Workflow

1. Text / image / multi-image (SEEDANCE_2_0, ≤3).
2. Collect optional wording tweaks; default to template lines if user says “use default.”
3. Expand → validate → confirm → wait.
4. [Video](url) only for links.

CLI reference

node {baseDir}/scripts/video_gen.js wait --json '{"model":"KLING_V3_0_PRO","prompt":"…","duration":10,"aspect_ratio":"9:16","generate_audio":true,"negative_prompt":"illegible text, watermark, garbled captions"}'
node {baseDir}/scripts/video_gen.js wait --json '{"model":"SEEDANCE_2_0","prompt":"…","duration":10,"aspect_ratio":"9:16","resolution":"720p","generate_audio":true}'

Full reference: WERYAI_VIDEO_API.md.

Definition of done

Video or error; five timed English lines present in prompt, with numeric second ranges matching the chosen duration.

Boundaries (out of scope)

• Not factual platform internals documentation—rhetorical short only.
• resources/WERYAI_VIDEO_API.md for CLI; no negative_prompt on Seedance.
• Multi-image only SEEDANCE_2_0, max 3.

Example prompts

• Default five-beat feed paranoia, cyberpunk UI
• Softer aesthetic: pastel social app, same five captions
• Screenshot-style image URL → animate with same subtitle schedule
• Parental controls aesthetic but same five-beat ‘who’s really in control’ lines

Model and API constraints (frozen for this skill)

Text-to-video

Image-to-video

Recommended models

Default parameters

Text-to-video (primary)

Default five-line schedule unless user overrides with a written table.

Image-to-video (optional)

Phone UI still → animate with parallax + same captions.

Tips

• First 2s should sync a hard cut to the first caption.
• Shorter words survive model text rendering better.

Packaged as feed-designs-you-video-gen.