Eye Roll Expression Video · skill pública de Open Skill
Eye Roll Expression Video
Eye Roll Expression Video es una Open Skill pública en Digen. Ábrela para ver qué crea este flujo.
Sports
1 0 3
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4284863/2026-10-19/z7bwxgbe.jpg
--- allowed-tools: - recognize_image - generate_image - generate_video --- # Eye Roll Expression Video Turn one uploaded reference image into a short photorealistic clip whose single visual idea is an eye-roll expression. The subject, identity and framing come from the user's image; the app only adds the attitude: eyes rolling up until mostly the whites show, lids half lowered, bored unimpressed face. There is no structured project state. Work in the conversation, keep the prompt you used in your reply, and reference canvas assets instead of pasting URLs. ## When to use - User uploads an image and asks for a short video, clip, animation or "make her roll her eyes". - User wants the same eye-roll clip redone with more/less attitude, different pacing, longer or shorter, or a different aspect ratio. - User asks for a still frame of the eye-roll expression only (then skip all video steps). ## Steps 1. **Read the reference image.** Call `recognize_image` on the uploaded image with a prompt asking for: subject age/look, hair, eye color and gaze direction, clothing, foreground props or hands, background, lighting, color palette, camera angle, framing/crop. Save these notes — they are the identity lock for every later prompt. 2. **Load the visual rules.** Call `read_skill_file` with `path="references/style-and-prompt.md"` and follow it. It defines Style, Subject/content, Layout, Must keep, Must avoid and the copyable English prompt template. Fill the template with the notes from step 1; do not write a generic prompt. 3. **Load the video rules and the script spec.** Call `read_skill_file` with `path="references/video-content-constraints.md"`, then `read_skill_file` with `path="references/video-script.md"`. 4. **Write the script for this request.** Following `references/video-script.md`, write a complete English script for this specific user request (logline, duration, beats with shot size and camera move, eyelid/eyeball action, head motion, light and atmosphere, transition, must-keep, must-avoid). Use the user's own words for mood, pace and length; use the step 1 notes for identity. A reusable boilerplate script is not acceptable. 5. **Decide the start frame** (branch rules live in `references/step-details.md`). If the uploaded image already reads as a tight, front-facing face shot in the app's soft pink/cream look, use it directly as the first frame. If it is wide, low-res, heavily stylized, or the eyes are already rolled, first call `generate_image` with `model='i2i.hd.lite'`, `input_urls=[uploaded image]`, `orientation='portrait'`, a prompt built from `references/style-and-prompt.md` that keeps the identity and reframes to a close-up with neutral forward gaze, `name="Eye Roll Start Frame"`, `entity_type='storyboard_images'`. That output becomes the first frame. 6. **Generate the video.** Compress the script into one English video prompt (action + camera + light + palette + must-keep) and call `generate_video` with: - `input_urls=[first frame asset]` (the start frame from step 5), - `model='i2v'`, - `prompt=<compressed script prompt>`, - `orientation='portrait'` (default for face close-ups; use `landscape` only if the user asks for widescreen or the source image is clearly wide), - `resolution='1080P'`, - `duration` short — 4 to 6 seconds unless the user asks for longer, - `name="Eye Roll Expression"`, - `entity_type='final_videos'`. If the user only wants a still image, call `generate_image` with `model='i2i.hd.lite'` instead and stop here. 7. **Report.** In natural language say what was kept from their image, what the expression does, and the length/ratio. Do not paste long URLs. In the same turn call `set_guidance` with 2–4 relevant next actions (e.g. "make the eye roll slower", "add a small head tilt", "regenerate with softer light", "give me a still frame"). ## Notes - Identity first: never change face, hair, clothing or signature props. The only things that move are eyes, lids, brows,